WO2023074542A1 - 合成材料選択方法、材料製造方法、合成材料選択データ構造、及び製造方法 - Google Patents
合成材料選択方法、材料製造方法、合成材料選択データ構造、及び製造方法 Download PDFInfo
- Publication number
- WO2023074542A1 WO2023074542A1 PCT/JP2022/039200 JP2022039200W WO2023074542A1 WO 2023074542 A1 WO2023074542 A1 WO 2023074542A1 JP 2022039200 W JP2022039200 W JP 2022039200W WO 2023074542 A1 WO2023074542 A1 WO 2023074542A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- feasibility
- synthesis
- physical property
- information
- database
- Prior art date
Links
- 239000000463 material Substances 0.000 title claims abstract description 803
- 230000015572 biosynthetic process Effects 0.000 title claims abstract description 333
- 238000003786 synthesis reaction Methods 0.000 title claims abstract description 333
- 238000010187 selection method Methods 0.000 title claims abstract description 72
- 238000004519 manufacturing process Methods 0.000 title claims description 305
- 230000000704 physical effect Effects 0.000 claims abstract description 319
- 238000000034 method Methods 0.000 claims abstract description 199
- 230000008569 process Effects 0.000 claims abstract description 140
- 238000012545 processing Methods 0.000 claims description 222
- 238000005259 measurement Methods 0.000 claims description 168
- 229920002994 synthetic fiber Polymers 0.000 claims description 97
- 230000002194 synthesizing effect Effects 0.000 claims description 78
- 238000003860 storage Methods 0.000 claims description 39
- 230000010365 information processing Effects 0.000 claims description 35
- 238000010801 machine learning Methods 0.000 claims description 32
- 230000004044 response Effects 0.000 claims description 11
- 239000002131 composite material Substances 0.000 claims description 2
- 239000003795 chemical substances by application Substances 0.000 description 37
- 238000004458 analytical method Methods 0.000 description 32
- 238000004891 communication Methods 0.000 description 32
- 238000010586 diagram Methods 0.000 description 30
- 230000006870 function Effects 0.000 description 30
- 238000006243 chemical reaction Methods 0.000 description 20
- 238000004422 calculation algorithm Methods 0.000 description 13
- 230000000052 comparative effect Effects 0.000 description 13
- 108090000765 processed proteins & peptides Proteins 0.000 description 10
- 239000002904 solvent Substances 0.000 description 10
- 239000010409 thin film Substances 0.000 description 10
- 102000004196 processed proteins & peptides Human genes 0.000 description 9
- 230000035899 viability Effects 0.000 description 9
- 238000000862 absorption spectrum Methods 0.000 description 8
- 230000003993 interaction Effects 0.000 description 8
- 238000011156 evaluation Methods 0.000 description 7
- 230000007246 mechanism Effects 0.000 description 7
- 239000000126 substance Substances 0.000 description 7
- 102000053602 DNA Human genes 0.000 description 6
- 108020004414 DNA Proteins 0.000 description 6
- 238000000231 atomic layer deposition Methods 0.000 description 6
- 239000013078 crystal Substances 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 6
- 229910010272 inorganic material Inorganic materials 0.000 description 6
- 239000011147 inorganic material Substances 0.000 description 6
- 230000003287 optical effect Effects 0.000 description 6
- 239000012071 phase Substances 0.000 description 6
- 230000002159 abnormal effect Effects 0.000 description 5
- 150000001413 amino acids Chemical class 0.000 description 5
- 239000012620 biological material Substances 0.000 description 5
- 238000000975 co-precipitation Methods 0.000 description 5
- 238000000576 coating method Methods 0.000 description 5
- 150000001875 compounds Chemical class 0.000 description 5
- 238000001514 detection method Methods 0.000 description 5
- 239000000975 dye Substances 0.000 description 5
- 238000001704 evaporation Methods 0.000 description 5
- 230000008020 evaporation Effects 0.000 description 5
- 238000005194 fractionation Methods 0.000 description 5
- 239000008204 material by function Substances 0.000 description 5
- 238000007578 melt-quenching technique Methods 0.000 description 5
- 239000002184 metal Substances 0.000 description 5
- 239000000203 mixture Substances 0.000 description 5
- 239000002086 nanomaterial Substances 0.000 description 5
- 239000002105 nanoparticle Substances 0.000 description 5
- 229920001184 polypeptide Polymers 0.000 description 5
- 108090000623 proteins and genes Proteins 0.000 description 5
- 102000004169 proteins and genes Human genes 0.000 description 5
- 230000006798 recombination Effects 0.000 description 5
- 238000003980 solgel method Methods 0.000 description 5
- 238000010532 solid phase synthesis reaction Methods 0.000 description 5
- 238000004544 sputter deposition Methods 0.000 description 5
- 235000000346 sugar Nutrition 0.000 description 5
- 239000013077 target material Substances 0.000 description 5
- 241000894006 Bacteria Species 0.000 description 4
- 230000005875 antibody response Effects 0.000 description 4
- 238000010170 biological method Methods 0.000 description 4
- 238000009795 derivation Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 238000001727 in vivo Methods 0.000 description 4
- 150000008163 sugars Chemical class 0.000 description 4
- 230000001988 toxicity Effects 0.000 description 4
- 231100000419 toxicity Toxicity 0.000 description 4
- 238000013528 artificial neural network Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 3
- 238000013461 design Methods 0.000 description 3
- 230000003834 intracellular effect Effects 0.000 description 3
- 229920002521 macromolecule Polymers 0.000 description 3
- 230000015654 memory Effects 0.000 description 3
- 238000005457 optimization Methods 0.000 description 3
- 238000004088 simulation Methods 0.000 description 3
- 238000012706 support-vector machine Methods 0.000 description 3
- 238000012546 transfer Methods 0.000 description 3
- 210000001835 viscera Anatomy 0.000 description 3
- 101100261173 Arabidopsis thaliana TPS7 gene Proteins 0.000 description 2
- GLQOALGKMKUSBF-UHFFFAOYSA-N [amino(diphenyl)silyl]benzene Chemical compound C=1C=CC=CC=1[Si](C=1C=CC=CC=1)(N)C1=CC=CC=C1 GLQOALGKMKUSBF-UHFFFAOYSA-N 0.000 description 2
- 230000006399 behavior Effects 0.000 description 2
- 239000003153 chemical reaction reagent Substances 0.000 description 2
- 238000004590 computer program Methods 0.000 description 2
- 238000007796 conventional method Methods 0.000 description 2
- 238000004090 dissolution Methods 0.000 description 2
- 239000012634 fragment Substances 0.000 description 2
- 238000001948 isotopic labelling Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000000329 molecular dynamics simulation Methods 0.000 description 2
- 230000037361 pathway Effects 0.000 description 2
- 238000002360 preparation method Methods 0.000 description 2
- 238000010845 search algorithm Methods 0.000 description 2
- 239000004065 semiconductor Substances 0.000 description 2
- 241000894007 species Species 0.000 description 2
- 238000012901 QM/MM calculation Methods 0.000 description 1
- 230000001580 bacterial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000013523 data management Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 239000003925 fat Substances 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- -1 low molecules Substances 0.000 description 1
- 230000014759 maintenance of location Effects 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000002156 mixing Methods 0.000 description 1
- 210000000056 organ Anatomy 0.000 description 1
- 230000001151 other effect Effects 0.000 description 1
- 239000000049 pigment Substances 0.000 description 1
- 238000009789 rate limiting process Methods 0.000 description 1
- 238000000611 regression analysis Methods 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 150000003384 small molecules Chemical class 0.000 description 1
- 238000001179 sorption measurement Methods 0.000 description 1
- 230000008685 targeting Effects 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 239000013598 vector Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
- G16C20/00—Chemoinformatics, i.e. ICT specially adapted for the handling of physicochemical or structural data of chemical particles, elements, compounds or mixtures
- G16C20/10—Analysis or design of chemical reactions, syntheses or processes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/23—Updating
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
- G16C20/00—Chemoinformatics, i.e. ICT specially adapted for the handling of physicochemical or structural data of chemical particles, elements, compounds or mixtures
- G16C20/30—Prediction of properties of chemical compounds, compositions or mixtures
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
- G16C20/00—Chemoinformatics, i.e. ICT specially adapted for the handling of physicochemical or structural data of chemical particles, elements, compounds or mixtures
- G16C20/70—Machine learning, data mining or chemometrics
Definitions
- the present disclosure relates to synthetic material selection methods, material manufacturing methods, synthetic material selection data structures, and manufacturing methods.
- Patent Document 1 a method of searching for a material to be newly synthesized by searching for a combination of physical property parameters of materials.
- the present disclosure proposes a synthesis material selection method, a material manufacturing method, a synthesis material selection data structure, and a manufacturing method that can enable material synthesis based on synthesis feasibility.
- one aspect of the synthesis material selection method performs material selection for selecting a material to be synthesized based on material property information and feasibility information in a database, In response, instructions are given for synthesis processing of the selected materials, and feasibility information in the database is updated based on the synthesis processing result from the control device.
- FIG. 1 is a diagram showing a configuration example of a material searching and manufacturing system of the present disclosure
- FIG. It is a flowchart which shows the processing procedure by a material search manufacturing system.
- Fig. 4 is a flow chart showing a process procedure for feasibility; 1 is a diagram showing an example of a physical property database according to an embodiment of the present disclosure;
- FIG. 2 illustrates an example of a viability database according to an embodiment of the present disclosure;
- FIG. FIG. 4 is a diagram showing an example of a material table according to an embodiment of the present disclosure;
- FIG. FIG. 4 is a diagram showing an example of prediction of physical properties according to an embodiment of the present disclosure;
- FIG. 2 illustrates an example of a feasibility assessment according to embodiments of the present disclosure
- FIG. 10 is a diagram showing an example of a search using a physical property prediction model and a feasibility model
- FIG. 11 shows a detailed example of a search using a physical property prediction model and a feasibility model
- FIG. 11 shows a detailed example of a search using a physical property prediction model and a feasibility model
- FIG. 11 shows a detailed example of a search using a physical property prediction model and a feasibility model
- FIG. 11 shows a detailed example of a search using a physical property prediction model and a feasibility model
- FIG. 11 shows a detailed example of a search using a physical property prediction model and a feasibility model
- FIG. 11 shows a detailed example of a search using a physical property prediction model and a feasibility model
- FIG. 11 shows a detailed example of a search using a physical property prediction model and a feasibility model
- FIG. 4 is a diagram showing an example of a viability database
- FIG. 4 is a diagram showing an example of a viability database
- FIG. 4 is a diagram showing an example of a viability database
- FIG. 4 is a diagram showing an example of a viability database
- FIG. 4 is a diagram showing an example of a viability database
- FIG. 4 is a sequence diagram showing an example of a processing procedure in the material searching and manufacturing system
- FIG. 4 is a sequence diagram showing an example of a processing procedure in the material searching and manufacturing system; It is a figure which shows an example of a structure of a material search manufacturing system. It is a figure which shows the structural example of the information processing apparatus of this indication. It is a figure which shows an example of a structure of a material manufacturing apparatus. It is a figure which shows an example of a structure of a material manufacturing apparatus. It is a figure which shows an example of a structure of a material manufacturing apparatus. 1 is a hardware configuration diagram showing an example of a computer that implements functions of an information processing apparatus; FIG. It is a figure which shows an example of the material manufacture result regarding a search. It is a figure which shows an example of the material manufacture result regarding a search.
- Embodiment 1-1 Outline of Configuration and Processing of Material Exploring and Manufacturing System According to Embodiment of Present Disclosure
- 1-1-1 Device configuration of material searching and manufacturing system 1-1-2. Overall processing flow of material searching and manufacturing system 1-2. Components of material searching and manufacturing system 1-2-1. Material manufacturing equipment 1-2-2. Control system 1-2-2-1. Data 1-2-2-2.
- Model 1-3 Example of database and model update processing after search and material manufacturing processing 1-3-1. Search using only physical property prediction model and update of database and model after material manufacturing process 1-3-2. Search using physical property prediction model and feasibility model and update of database and model after material manufacturing process 1-4. Details of database and model update processing after searching and material manufacturing processing 1-4-1.
- FIG. 1 is a diagram showing a configuration example of a material searching and manufacturing system of the present disclosure.
- the material searching and manufacturing system 1 includes a material manufacturing device 10 , a control device 20 and an external storage device 50 .
- the material manufacturing apparatus 10 and the control apparatus 20 are communicably connected by wire or wirelessly via a predetermined communication network (the control network N in FIG. 1).
- the control device 20 and the material manufacturing device 10 can exchange information with each other through the control network N.
- the control network N may be a dedicated network capable of communicating wirelessly or via wires, a local area network, a wide area network called the Internet or a cloud.
- the material manufacturing apparatus 10 has a synthesizing apparatus 11 that synthesizes materials, an analyzing apparatus 12 that identifies the synthesized materials, and a measuring apparatus 13 that evaluates physical properties.
- Each component (each device) of the material manufacturing device 10 such as the synthesizing device 11, the analyzing device 12, and the measuring device 13 can transfer materials to each other, and the configuration thereof will be described later.
- the external storage device 50 is a server device that holds information used for processing in the material searching and manufacturing system 1 . For example, the external storage device 50 provides information to the control device 20 .
- the control device 20 includes a control interface 21 for controlling the material manufacturing device 10, a physical property database 22 (physical property DB in FIG. 1) holding physical property data, and a feasibility database 23 holding whether synthesis and/or measurement could be executed. (Feasibility DB in FIG.
- search agent 26 that is a computer program (hereinafter , search agent 26), and a processing unit 202 (processor (CPU, GPU, GPGPU, neural network accelerator, or a combination of these, etc.), an operation unit 203 consisting of a user interface (keyboard, mouse, touch panel, microphone (voice input), camera (image input such as face)) for operating the control device, or A display unit 204 (such as a display) for displaying the processing result of the processing unit 202 is included.
- processor CPU, GPU, GPGPU, neural network accelerator, or a combination of these, etc.
- an operation unit 203 consisting of a user interface (keyboard, mouse, touch panel, microphone (voice input), camera (image input such as face) for operating the control device, or
- a display unit 204 (such as a display) for displaying the processing result of the processing unit 202 is included.
- the physical property prediction model 24 and the feasibility model 25 may be stored in the same storage unit 201 as the physical property database 22 (physical property DB in FIG. 1) or the feasibility database 23 (feasibility DB in FIG. 1). Alternatively, it may be stored in another storage device.
- the physical property prediction model 24, the feasibility model 25, or the search agent 26 may also be stored in the same storage unit 201 as the physical property prediction model 24 and the feasibility model 25, or may be stored in a separate storage device.
- part of the storage unit 201 may be configured as a device (for example, the external storage device 50 or the like) separate from the control device 20 and connected to the control network N via a network interface (not shown). Further, the storage unit 201 and another storage device (for example, the external storage device 50 or the like) may cooperate with each other to form a virtual storage using functions of an operating system (not shown).
- communication control devices/communication infrastructure such as routers and switches are used between the control interface 21 and the control network N, and between the control network N and the synthesizing device 11, the analyzing device 12, or the measuring device 13. data communication may be performed.
- the search agent 26 may implement its functions as a computer program, or may implement its functions by means of hardware, such as a dedicated processor (not shown). It may work. Furthermore, as will be described later, the physical property prediction model 24 and the feasibility model 25 are updated by re-learning.
- the physical property prediction model 24 and the feasibility model 25 may be machine learning models realized by machine learning algorithms such as Bayesian optimization, neural networks, and SVM (Support Vector Machine).
- the storage unit 201 further stores a software program for executing a machine learning algorithm (not shown).
- the physical property prediction model 24 is read from the storage unit 201 and shown in the figure.
- the physical property prediction model 24 can be updated using machine-learned data that does not apply.
- the processing unit 202 copies the physical property prediction model 24 in the process of being updated to the physical property prediction model update area 24', and generates an updated model of the physical property prediction model 24.
- an updated physical property prediction model can be generated based on the physical property prediction model 24. That is, the processing unit 202 copies the physical property prediction model 24, which is a machine learning model read from the storage unit 201, to the physical property prediction model update area 24' using a software program that executes a machine learning algorithm. By applying the machine learning data, an updated machine learning model can be produced. By writing the manufactured updated machine learning model again to the storage unit 201 as the physical property prediction model 24, the updated machine learning model can be continuously used as the physical property prediction model 24.
- the control device 20 can perform processing for manufacturing the physical property prediction model 24, which is an updated machine learning model.
- generation of an updated feasibility model can be performed.
- the feasibility model update area 25' of the storage unit 201 is used. That is, the processing unit 202 copies the feasibility model 25, which is the machine learning model read from the storage unit 201, to the feasibility model update area 25' using a software program that executes a machine learning algorithm. By applying machine learning data to it, it is possible to manufacture an updated machine learning model. By writing the manufactured updated machine learning model again to the storage unit 201 as the feasibility model 25, the updated machine learning model can be continuously used as the feasibility model 25. .
- control device 20 can perform processing for manufacturing the feasibility model 25, which is the updated machine learning model.
- the display unit 204 can display data structures illustrated in FIGS. 4 to 20 according to the input from the operation unit 203 .
- processing unit 202 can issue an instruction to start, stop, or modify the processing shown in FIG.
- FIG. 2 is a flow chart showing a processing procedure by the material searching and manufacturing system.
- FIG. 2 shows an example of a material manufacturing method performed by the material searching and manufacturing system 1 .
- the material searching and manufacturing system 1 is driven by the searching agent 26 and searches for materials according to the following procedure.
- the material searching and manufacturing system 1 sets the range of the material searching space and the target values of the material physical properties (step S1). Then, the material searching and manufacturing system 1 sets a material searching space (step S2). For example, the material search and manufacturing system 1 uses prediction results from the physical property prediction model 24 and/or the prediction results from the feasibility model 25 to prioritize materials included in the material search space.
- the prediction result (output) of the physical property prediction model 24 or the feasibility model 25 may be regression analysis data, which will be described later in detail.
- the physical property prediction model 24 and the feasibility model 25 may be machine learning models using machine learning algorithms such as Bayesian optimization, neural networks, and SVM. Details of this will be described later.
- the material searching and manufacturing system 1 selects candidate materials based on the priority of the materials (step S3).
- the control device 20 of the material searching and manufacturing system 1 selects a material to be synthesized based on material property information in the property database 22 and feasibility information in the feasibility database 23 .
- the material searching and manufacturing system 1 synthesizes candidate materials (step S4).
- the control device 20 of the material searching and manufacturing system 1 instructs the material manufacturing device 10 of the material searching and manufacturing system 1 to synthesize the selected materials, and the material manufacturing device 10 responds to the instruction from the control device 20 Synthesize the materials accordingly.
- step S5 If the synthesis fails (step S5: No), the material searching and manufacturing system 1 updates the data in the item indicating "feasibility" of the feasibility database 23 to indicate that the synthesis has "failed", and further The feasibility model 25 is updated so that the "predicted value" of the "feasibility” of the material for which synthesis failed (step S9), and the process returns to step S2. repeat.
- the material manufacturing device 10 of the material searching and manufacturing system 1 transmits the result of synthesis processing to the control device 20 of the material searching and manufacturing system 1, and the control device 20, based on the result of the synthesis processing from the material manufacturing device 10, Update the feasibility information in the feasibility database 23 .
- the predicted value of "feasibility" under the same synthesis conditions is By making it lower, in the selection of candidate materials from the next time onwards, the priority will be lowered under the same conditions as the synthesis failure, and the probability that materials other than the failed synthesis material will be selected will be increased. can be done.
- step S6 the material searching and manufacturing system 1 measures the physical properties of the material.
- step S7 When the physical property measurement fails (step S7: No), the material searching and manufacturing system 1 indicates that the synthesis was “successful” but the measurement was “failed” in the item indicating "feasibility” of the feasibility database 23.
- the data are updated as shown, and the feasibility model 25 is further updated so that the "predicted value" of the "feasibility” for the successfully synthesized material is lower in the conditions where the measurement failed (step S9), return to step S2 and repeat the process.
- the predicted value of "feasibility" under the same synthesis conditions for the material whose measurement failed is By making it lower, in the selection of candidate materials from the next time onwards, the priority will be lowered under the same conditions as the physical property measurement failure, and the probability of selecting a material other than the material whose synthesis failed will be increased. be able to.
- step S7 the material searching and manufacturing system 1 determines whether the physical property value has reached the target value (step S8).
- step S8 the material searching and manufacturing system 1 updates the "physical property" of the physical property database 22 with the measured physical property value, and in the physical property prediction model 24, synthesis Update is performed so that the measured physical property value is predicted to be lower than the target value under the same synthesis conditions for the material for which the synthesis was successful (step S10).
- step S10 the item indicating "feasibility" of the feasibility database 23
- data is updated to indicate that both synthesis and measurement were “successful”
- the “predicted value” of the "feasibility” is updated so as to be higher under the condition that the synthesis is successful (step S9), and the process returns to step S2 to repeat the process.
- the predicted value of "physical property" under the same synthesis conditions is By making it lower than the target value, in the selection of candidate materials from the next time onwards, the priority will be lowered under the same conditions as those that failed to achieve the target value, and materials other than the materials that failed to achieve the target value material is selected.
- the material searching and manufacturing system 1 updates the "physical property" of the physical property database 22 with the measured physical property value, and the physical property prediction model 24 succeeds in synthesis. Update is performed so that the measured physical property values are predicted under the same synthesis conditions for the materials obtained (step S11).
- the item indicating "feasibility" in the feasibility database 23 is updated to data indicating that both the synthesis and the measurement were “successful”, and further, in the feasibility model 25, the material that was successfully synthesized is "executed
- the feasibility model is updated so that the "predicted value” of "possibility” is 1 or a value as close to 1 as possible (step S12), and the material search is terminated.
- the predicted value of "feasibility” will be 1 in the prediction for materials that have been synthesized successfully in the next material search. Or, by checking with the "synthesized material list" before prediction, the priority will be lowered in the selection of candidate materials from the next time onwards, and materials that have already been successfully synthesized will not be selected.
- the prediction results based on the physical property prediction model 24 are referred to as "priority”
- the prediction results based on the feasibility model 25 are referred to as "feasibility”. ", but it should be understood that these are one of the embodiments.
- the selection order determined based on the prediction results of these two models is collectively referred to as "priority" as a higher concept.
- materials search In conventional materials search technology, materials search generally uses high-precision first-principles calculations, QM/MM calculations, molecular dynamics methods, coarse-grained molecular dynamics methods, multi-scale simulations, and other computationally expensive methods. was using In the material search based on such computational theory, the feasibility of material manufacturing (synthesis feasibility, measurability) was not considered, so materials that could not actually be synthesized or measured were selected as candidate materials. , failed to synthesize and evaluate using selected materials, and could not select suitable materials. In this way, the conventional techniques for searching for materials do not consider the feasibility (synthesis feasibility, measurability) in material manufacturing, so materials that cannot actually be synthesized or measured have been used as candidate materials. , search efficiency was poor.
- the material search in the material searching and manufacturing system 1 since the search is performed taking into consideration the feasibility of synthesis, it is possible to synthesize materials based on the feasibility of synthesis.
- the material search in the material search and manufacturing system 1 by updating the feasibility model 25 based on the synthesis and measurement processing, by considering the possibility of success, the prediction accuracy is increased, and the number of executions is small. New materials with excellent properties (physical properties) can be discovered.
- the material to be synthesized in consideration of the feasibility of synthesis even when searching for materials in the set search space, conventionally, it is only necessary to have no experience in synthesizing. By allowing materials that have been excluded from options to be preferentially selected as synthesis candidates, it is possible to first give opportunities for synthesis processing to materials that would not have been explored in the past. It is possible to increase the possibility of searching for new materials.
- the material searching and manufacturing system 1 is a material searching and manufacturing system that incorporates an automatic material manufacturing device, a physical property prediction model, and a feasibility (synthesis possibility, measurability) model. Therefore, the material search and manufacturing system 1 can realize highly efficient search and synthesis of materials, confirm the feasibility of synthesizing materials, and quickly obtain a design space for materials that can be synthesized and measured, thereby improving search efficiency. can be made
- the synthesizer 11 internally has a mixing/reaction vessel, reagent/solvent stock, and product separation function, or is connected to an external device having these functions, and these components can exchange contents with each other.
- the synthesizer 11 also mixes reagents and solvents, synthesizes candidate materials, and separates products based on commands from the controller 20 . Further, the synthesizer 11 determines whether or not these operations are successful using a detection unit (a detection circuit or a detection software program running on a processor; the same shall apply hereinafter), and reports the determination result to the control device 20 ( Send.
- a detection unit a detection circuit or a detection software program running on a processor; the same shall apply hereinafter
- the analysis device 12 receives the product synthesized by the synthesis device 11 and determines whether the product matches the candidate material. Thereby, the analysis device 12 generates an analysis result.
- the analysis device 12 also has a detection unit that determines whether the analysis is successful (generates a success determination result), and transmits the analysis result and the success determination result to the control device 20 .
- the measuring device 13 receives the material synthesized by the synthesizing device 11 and determined by the analyzing device 12 to match the candidate material, and measures the physical properties of the material. Thereby, the measuring device 13 generates a measurement result.
- the measurement device 13 has a detection unit that determines whether the measurement of physical properties has succeeded (generates a success determination result), and transmits the measurement result and the success determination result to the control device 20 .
- the material manufacturing device 10 manufactures materials.
- the material manufacturing apparatus 10 manufactures materials under the control of the control device 20 .
- the material manufacturing apparatus 10 receives an instruction to synthesize materials selected using the material property information and the feasibility information
- the material manufacturing apparatus 10 manufactures materials according to the instruction to synthesize the materials.
- the material manufacturing apparatus 10 performs material manufacturing by the synthesizing apparatus 11 for the material indicated by the synthesizing process instruction.
- the material manufacturing apparatus 10 may be configured to cooperate with a plurality of different material manufacturing apparatuses 10 in data sharing and processing control through a communication network. It suffices if it has at least one function of analysis or measurement processing.
- the material manufacturing apparatus 10 may internally have software having a physical property evaluation function for evaluating physical properties, or may be used by connecting to an external device having this function. In that case, the material manufacturing apparatus 10 executes material simulation according to an instruction from the control device 20 . The material manufacturing apparatus 10 determines whether the material synthesis has succeeded (generates a success determination result) and transmits the success determination result and the material synthesis result to the control device 20 .
- the material manufacturing apparatus 10 may be configured by hardware or may be configured by software. Moreover, there may be a plurality of material manufacturing apparatuses 10, and they may be connected via a predetermined network such as the Internet, but this point will be described later.
- the physical property prediction model 24 is a regression model that is estimated from data in the physical property database 22 and predicts physical property values from materials.
- a regression model may return only the predicted values for the material, or it may return the predicted values and the variance of the predicted values.
- the feasibility model 25 is represented by the following formula (1).
- F SYNTH is the Synthesizability Model (range 0-1)
- F MEAS is the Measurability Model (range 0) -1).
- the synthesizing possibility model F SYNTH is a model for predicting whether or not synthesis will succeed, and is represented by the following equation (2).
- F PATH is the Synthetic Path Model
- F Step is the Synthetic Step Model
- NS is the number of synthetic steps.
- the synthetic pathway model F PATH is a model indicating whether or not a synthetic pathway exists, and is represented by the following formula (3).
- the synthetic route model F PATH has a value of 0 or 1.
- the synthetic route model F PATH is 1 if there is a synthetic route, and 0 otherwise.
- the number of synthesis steps N S satisfies the following equation (4).
- N MAX indicates the maximum number of synthesis steps.
- the step synthesis possibility model F Step is a model for predicting whether or not one step synthesis process will succeed, and is represented by the following equation (5).
- the following term (6) is the step synthesis yield model (value range 0-1).
- the following term (7) is a product identification model (value range 0-1).
- the controller 20 predicts the yield of each synthesis step, predicts whether it can be determined that the product matches the target material, and predicts whether the product can be fractionated.
- the evaluability model F MEAS is a model that predicts whether a measurement will be successful.
- FIG. 3 corresponds to the flow of processing by the feasibility model 25 described above.
- FIG. 3 is a flow chart illustrating the process for feasibility.
- the feasibility model 25 derives each value in each process corresponding to steps S21 to S24 in FIG. 3, and generates an output value (score) indicating feasibility based on the value.
- the feasibility model 25 derives a value according to the synthetic route model F PATH (step S21).
- the feasibility model 25 derives a value according to the step synthesis feasibility model F Step according to the number of steps (steps S22 and S23).
- the feasibility model 25 derives values according to the step synthesizing feasibility model F Step corresponding to each of the first to Ns-th synthesizing steps. . Also, the feasibility model 25 derives a value based on the evaluability model F MEAS (step S24). Then, the feasibility model 25 derives the value of the feasibility model F FS using the values of the synthesis path model F PATH , the step synthesis feasibility model F Step , the evaluability model F MEAS and the like.
- the model group included in the feasibility model 25 may be a model estimated from values obtained in the material manufacturing process or simulation results, or may be represented by a constant or a simplified empirical formula. may be used.
- the search agent 26 determines the priority of candidate materials based on the predicted value obtained from the physical property prediction model 24 and/or the value obtained from the feasibility model 25, and according to this priority, the candidate to be synthesized and measured next. Select materials and perform synthesis and measurement of selected candidate materials. Synthesis and measurement are performed by issuing commands from the search agent 26 to the material manufacturing apparatus 10 through the control interface 21 .
- the control device 20 receives "whether it could be executed" (feasibility data) and "measured values” (physical property data) reported from the material manufacturing apparatus 10 through the control interface 21, and stores the feasibility database 23 and the feasibility model, respectively. 25, updating the physical property database 22 and the physical property prediction model 24;
- the search agent 26 may reside on the same computing device as the control interface 21 or on a separate computing device.
- the control device 20 selects materials using the material property information stored in the property database 22 and the feasibility information stored in the feasibility database 23. For example, the control device 20 instructs the material manufacturing device 10 to synthesize the selected materials. For example, the control device 20 updates the feasibility information stored in the feasibility database 23 based on the synthesizing processing result from the material manufacturing device 10 .
- FIG. 4 is a diagram showing an example of a physical property database according to an embodiment of the present disclosure
- a database DB1 shown in FIG. 4 corresponds to the physical property database 22 .
- the database DB1 stores data (learning data) used for learning the physical property prediction model 24 .
- the database DB1 includes items such as "ID”, "molecule ID”, “descriptor”, "physical property value", and "material attribute”.
- a database DB1 which is a physical property database, contains two or more materials to be synthesized.
- ID indicates identification information for identifying data.
- Molecule ID indicates identification information for identifying a molecule.
- Descriptor indicates a descriptor. The example in FIG. 4 shows a case where "LabuteASA”, “Chi0v”, “Chi0n”, “Chi0”, “Kappa1”, etc. are used as descriptors, but FIG. Any one can be used.
- the “descriptor” item stores data (values) that quantify the chemical characteristics of the corresponding molecule.
- Physical property value indicates the physical property value of the molecule.
- the example in FIG. 4 shows a case where "Heat Capacity” is used as a physical property value, but FIG. 4 is only an example, and any “physical property value” can be adopted.
- Physical properties include mechanical properties, thermal properties, electrical properties, magnetic properties, optical properties, electrochemical properties, efficacy, toxicity, antibody response, interaction with cells, and interaction with internal organs. , intracellular transportability, in vivo transportability, adsorptivity, and solubility.
- the item “physical property value” stores data (values) relating to properties possessed by the molecule.
- the value of the "physical property value” item is updated to the measured physical property value when the measurement of the molecule is successful. For example, the control device 20 updates the value of "physical property” based on the measurement result.
- “Material attribute” indicates the attribute of a molecule (material) as a material.
- “#1” storing one material attribute is shown for explanation, but "#2", “#3", etc. are provided as many as the number of material attributes.
- “Material attributes” are small molecules, dyes, macromolecules, fluorescent/identifying isotope labels, self-organizing materials/structures, biomaterials (sugars, peptides, polypeptides, amino acids, proteins, fatty compounds, DNA (Deoxyribonucleic Acid ), etc.), organic thin films (evaporation, coating processes), inorganic materials (solid-phase method, coprecipitation method, melt quenching method, sol-gel method, etc.), nanoparticles, metal complexes, inorganic thin films (ALD (atomic layer deposition), sputtering etc.), synthetic materials based on synthetic biological techniques (genetic recombination and material synthesis using bacteria), functional materials with crystal structures, nanostructures, and microstructures.
- the item "material attribute” stores information indicating the material attribute of the molecule (material).
- the information (material physical property information) stored in the database DB1, which is an example of the physical property database 22, is structured to include at least two items, one of which is the material is an attribute item of
- the database DB1 is not limited to the above, and may store various types of information depending on the purpose.
- the database DB1 stores information indicating molecules (materials) that have been successfully synthesized in the past.
- the database DB1 may store a flag (value) indicating a molecule (material) that has been successfully synthesized in the past in association with the molecule ID.
- Database DB1 may exclude molecules that have been successfully synthesized in the past.
- FIG. 5 is a diagram illustrating an example of the viability database 23 according to an embodiment of the present disclosure.
- a database DB2 shown in FIG. 5 corresponds to the feasibility database 23 .
- the database DB2 stores data (learning data) used for learning the feasibility model 25 .
- the database DB2 includes items such as "ID”, "molecule ID”, "descriptor”, and "executability”.
- ID indicates identification information for identifying data.
- Molecule ID indicates identification information for identifying a molecule.
- Descriptor indicates a descriptor. The example in FIG. 5 shows a case where "TPSA”, “qed”, “SlogP_VSA2”, “SlogP_VSA5", etc. are used as descriptors, but FIG. Adoptable.
- the “descriptor” item stores data (values) that quantify the chemical characteristics of the corresponding molecule.
- “Performance” indicates the viability of the molecule.
- FIG. 5 shows a case where "synthesis pass/fail” is used as an item related to feasibility, FIG. 5 is only an example, and any “feasibility” can be adopted. “Feasibility” may be various elements related to feasibility such as measurability.
- the “executability” item stores data (value) indicating whether or not the element is executable.
- the database DB2 may store various types of information, not limited to the above, depending on the purpose.
- the database DB2 stores information indicating molecules (materials) that have been successfully synthesized in the past.
- the database DB2 may store flags (values) indicating molecules (materials) that have been successfully synthesized in the past in association with their molecule IDs.
- the database DB2 may exclude molecules (materials) that have been successfully synthesized in the past. Examples of the feasibility database 23 other than the database DB2 will be described later.
- the data of the data records contained in the physical property database 22 and the feasibility database 23 may retain reference links so that they can be accessed mutually.
- agent software via relational database management software, may perform merge, filter operations, etc. between these two databases to provide integrated access to the data of any data record.
- what is written as a database in the present disclosure is assumed to include a plurality of data tables of the same type, and if there is a description suggesting a data table in the specification, it is managed in the database.
- the inclusion of the term table in a description of a database is intended to imply a broader concept of the database.
- FIG. 6 is a diagram illustrating an example of a material table according to an embodiment of the present disclosure.
- the control device 20 has a table DB3 shown in FIG.
- the table DB3 stores data (learning data) used for learning the feasibility model 25 .
- the table DB3 includes items such as "ID”, "molecule ID”, and "SMILES".
- ID indicates identification information for identifying data.
- Molecule ID indicates identification information for identifying a molecule.
- SMSILES indicates a representation of a molecule in SMILES notation. The “SMILES” item stores information in which the chemical structure of a molecule is converted into character strings using alphanumeric characters in ASCII code.
- material ID indicates identification information for identifying the material.
- Composition ratio is a column of numerical values indicating element species constituting the material and their content ratios.
- table DB3 may store various types of information, not limited to the above, depending on the purpose.
- FIG. 7 is a diagram showing an example of physical property prediction according to an embodiment of the present disclosure.
- the control device 20 derives a value corresponding to each descriptor using a character string in which the molecule identified by the molecule ID is expressed in SMILES notation. Then, the controller 20 inputs the values corresponding to the descriptors to the physical property prediction model 24, thereby causing the physical property prediction model 24 to output the predicted value of the physical property.
- the physical property prediction model 24 inputs the values corresponding to the descriptors "LabuteASA”, “Chi0v”, “Chi0n”, “Chi0", and “Kappa1", and the average predicted for the physical property "Heat Capacity” and variance are output as the predicted value of the physical property "Heat Capacity".
- the control device 20 outputs a predicted value of physical properties (hereinafter also referred to as a “first predicted value”) from the physical property prediction model 24 by inputting a material feature amount related to a molecule (material) to the physical property prediction model 24.
- the physical property prediction model 24 outputs a first prediction value in response to input of material feature amounts of two or more materials constituting the material.
- the material feature value here is, for example, a numerical sequence uniquely determined from the molecular structure, such as the molecular fingerprint (ECFP4) determined from the bonding relationship between atoms in the molecule, the Coulomb matrix, the number of atomic species contained, etc. molecular descriptors calculated from structural and structural features, molecular feature vectors extracted by a graph convolutional neural network (GCN), and the like.
- GCN graph convolutional neural network
- the physical property prediction model 24 may receive various information as input and output various information.
- the physical property prediction model 24 may output standard deviation instead of variance.
- the physical property prediction model 24 may receive a value corresponding to each descriptor as an input and output one value related to the physical property and a value indicating the degree of certainty of the value.
- the control device 20 may average one value related to the physical property and calculate the variance based on the confidence factor. For example, the control device 20 may calculate a smaller standard deviation as the degree of certainty is higher.
- a plurality of physical property prediction models 24 may be generated by learning for each physical property.
- the physical property prediction model 24 may be a model trained to output multiple physical properties.
- the controller 20 uses the data stored in the physical property database 22 as learning data to train and update the physical property prediction model 24 using a learning algorithm.
- the learning of the physical property prediction model 24 by the control device 20 is so-called supervised learning, and a detailed description thereof will be omitted.
- a physical property prediction model 24 may be learned.
- FIG. 8 is a diagram illustrating an example of a feasibility assessment according to an embodiment of the disclosure.
- the control device 20 derives a value corresponding to each descriptor using a character string in which the molecule identified by the molecule ID is expressed in SMILES notation. Then, the control device 20 inputs a value corresponding to each descriptor to the feasibility model 25 to obtain a predicted value of physical properties (hereinafter also referred to as a “second predicted value”) from the feasibility model 25. output.
- the feasibility model 25 outputs a second predicted value in response to input of material feature quantities of each of two or more materials that make up the material.
- the feasibility model 25 receives values corresponding to descriptors "TPSA”, “qed”, “SlogP_VSA2”, and “SlogP_VSA5" as inputs and outputs values indicating feasibility.
- the feasibility model 25 receives a value corresponding to each descriptor as an input and outputs a value between 0 and 1 indicating feasibility.
- the feasibility model 25 outputs a value that is closer to 1, indicating higher feasibility.
- the control device 20 inputs a material feature quantity related to a molecule (material) to the feasibility model 25 to output a value indicating feasibility from the feasibility model 25 .
- the control device 20 uses the data stored in the feasibility database 23 as learning data to train and update the feasibility model 25 with a learning algorithm.
- the learning of the feasibility model 25 by the control device 20 is so-called supervised learning, and a detailed description thereof will be omitted.
- the feasibility model 25 may be learned by a training process with a learning algorithm.
- FIG. 9 is a diagram showing an example of searching using a physical property prediction model. The processing shown in FIG. 9 corresponds to the processing flow of the flowchart shown in FIG.
- Each of the six tables corresponds to the state at the start of processing in each of the first to sixth times.
- Six search tables #1 in FIG. 9 show updates of the search tables #1 after the first to sixth processes (the processes in FIG. 2).
- the table arranged in the upper left row in FIG. 9 corresponds to the first search table #1.
- the table arranged in the upper center row in FIG. 9 corresponds to the search table #1 for the second time, that is, the search table #1 after the first process.
- Search table #1 has items such as "molecule ID”, "priority", and "synthesis success".
- "Molecule ID” indicates identification information for identifying a molecule.
- "priority” is a value indicating the priority.
- “priority” is a value (score) output by the physical property prediction model 24 in response to input of corresponding molecular information.
- “priority” is the first predicted value of the material based on material property information. The first predicted value is output from the physical property prediction model 24 in response to an input based on material feature amounts of each of two or more materials forming the material registered in the physical property database 22, for example.
- priority is assigned in descending order of the value of "priority” starting from the first place.
- the numerical value (such as “1") placed at the end of “priority” indicates the number of updates.
- “priority” without a numerical value at the end indicates “priority” before update, that is, at the start of processing.
- “priority 1” indicates the “priority” at the time of updating after the first processing. In this way, the number after “priority” is for convenience to indicate the number of updates, and "priority" and items with numbers at the end of “priority” are the same item “priority”.
- Synthesis success indicates whether the synthesis is successful. Note that FIG. 9 shows a state in which a value is stored in "synthesis success” to indicate where the synthesis is successful, but the "synthesis success” value of the molecule before synthesis processing may be unknown. Note that the search table #1 may include an item such as "measurement” that indicates the success or failure of the measurement process.
- the material searching and manufacturing system 1 performs the synthesis process on the molecule identified by the molecule ID "m100275", which has the highest priority.
- the synthesis of the molecule identified by the molecule ID "m100275” fails, and the material searching and manufacturing system 1 does not update the data.
- “success (DB update)” or “failure (DB unchanged)” is described together with white arrows between each search.
- DB means the physical property database 22 . 9 to 15, “success” means success of synthesis and success of measurement (“YES” in S7 of FIG. 2), and “failure” means success of synthesis (“NO” in S5 of FIG.
- the material searching and manufacturing system 1 performs the synthesis process, etc. on the molecule identified by the molecule ID "m059059", which has the fourth priority.
- FIG. 9 it is assumed that the molecule identified by the molecule ID "m059059” was synthesized and measured successfully in the fourth processing.
- the material searching and manufacturing system 1 updates the data relating to the physical property values of the molecule ID “m059059” with the fourth priority in the physical property database 22 .
- the material searching and manufacturing system 1 further re-learns (updates) the physical property prediction model 24 using the updated physical property database 22 . Then, the material searching and manufacturing system 1 updates the value of “priority” using the updated physical property prediction model 24 .
- “priority 1" after the table (search table #1 in the fifth search) arranged in the lower center row in FIG. 9 corresponds to the priority value derived using the physical property prediction model 24 after updating.
- the material searching and manufacturing system 1 performs the fifth and sixth processes in order from the first in order of priority based on the value of "priority1". For example, the material searching and manufacturing system 1 repeats the above-described processing until a molecule that satisfies desired physical property values is synthesized.
- FIG. 10 corresponds to the processing flow of the flowchart shown in FIG. 10 that are the same as those in FIG. 9 will be omitted as appropriate.
- search table #2 Each of the six tables (hereinafter also referred to as "search table #2") in FIG. ) at the start of processing in each of the first to sixth times.
- the six search tables #2 in FIG. 10 show updates of the search tables #2 after the first to sixth processes (the processes in FIG. 2).
- the table arranged in the upper left row in FIG. 10 corresponds to the first search table #2.
- the table arranged in the upper center row in FIG. 10 corresponds to the search table #2 for the second time, that is, the search table #2 after the first process.
- Search table #2 in FIG. 10 shows an example of the synthetic material selection data structure used in the material selection process.
- Search table #2 has items such as "molecule ID”, “priority”, “feasibility”, and “success”.
- "Molecule ID” and “priority” are the same as in FIG.
- "feasibility” is a value indicating feasibility.
- "feasibility” is a value (score) output by the feasibility model 25 in response to the input of corresponding molecular information.
- "feasibility” is the second predicted value of the material based on feasibility information. The second predicted value is output from the feasibility model 25 in response to an input based on material feature amounts of each of two or more materials forming the material registered in the feasibility database 23, for example.
- the material searching and manufacturing system 1 treats molecules whose “feasibility” value is equal to or greater than a predetermined threshold value (for example, 0.5).
- the number (such as "1") placed at the end of "feasibility” indicates the number of updates. For example, “feasibility” without a numerical value at the end indicates “feasibility” before update, that is, at the start of processing. Also, “feasibility 1" indicates “feasibility” at the time of one update. Thus, the number after “feasibility” is for convenience to indicate the number of updates, and "feasibility" and “feasibility” with a number at the end are the same item “feasibility”.
- “Success” as an item in the search table indicates whether synthesis and measurement are successful.
- FIG. 10 shows a state in which a value is stored in "success” to indicate where the process succeeds, but the value of "success” in the numerator before processing may be unknown.
- “success” or “failure” is indicated with a hollow arrow between each search.
- “success” means success in synthesis and success in measurement (“YES” in S7 of FIG. 2)
- “Failure” means failure of synthesis (“NO” in S5 of FIG. 2) or failure of measurement (“NO” in S7 of FIG. 2).
- the data is updated (physical property DB: update) regardless of whether the target value is achieved (S8 in FIG. 2).
- the data is not updated and is unchanged (physical property DB: unchanged).
- the execution database 23 is updated regardless of "success” or "failure” (executability DB: update).
- the molecules with the first to third priorities are the values output by the feasibility model 25, and the value of "feasibility” meaning feasibility is less than the threshold (0.5). Therefore, these molecules are not processed, and subsequent processing is skipped. In this way, the material searching and manufacturing system 1 uses the "feasibility" information regarding feasibility, so that unnecessary molecules are not subjected to processing, so that the processing efficiency can be improved.
- the material searching and manufacturing system 1 performs the synthesis process on the molecule identified by the molecule ID "m059059", which has the fourth priority and the value of "feasibility” is equal to or greater than the threshold value (0.5). etc.
- the material searching and manufacturing system 1 updates the data regarding the physical property values of the molecule ID "m059059” in the physical property database 22, and adds data regarding the feasibility of the molecule ID "m059059” in the feasibility model 25. (or update).
- the material searching and manufacturing system 1 further re-learns (updates) the physical property prediction model 24 using the updated physical property database 22 . Then, the material searching and manufacturing system 1 updates the value of “priority” using the updated physical property prediction model 24 .
- “priority 1" after the table (search table #2 in the second search) arranged in the upper center row in FIG. 10 corresponds to the priority value derived using the physical property prediction model 24 after updating.
- the material searching and manufacturing system 1 re-learns (updates) the feasibility model 25 using the feasibility database 23 after updating. Then, the material searching and manufacturing system 1 updates the value of “feasibility” using the updated feasibility model 25 . "feasibility 1" after the table (search table #2 in the second search) arranged in the upper center row in FIG.
- the material searching and manufacturing system 1 uses the molecule ID "m100686" which has the highest priority in the updated search table #2 and whose "feasibility" value is equal to or greater than the threshold value (0.5). Synthesis processing and the like are performed on the identified molecules.
- FIG. 10 it is assumed that the synthesis of the molecule identified by the molecule ID "m100686” has failed ("NO" in S7 of FIG. 2). In this case, the material searching and manufacturing system 1 adds (or updates) data on feasibility of the molecule ID “m100686” to the feasibility database 23 .
- the material searching and manufacturing system 1 adds (or updates) the information on the feasibility of the molecule identified by the molecule ID "m100686" to the feasibility database 23 as the information on the molecule whose synthesis failed.
- the database 23 is updated again, and the feasibility model 25 is re-learned (updated) using the updated feasibility database 23 .
- the material searching and manufacturing system 1 updates the value of “feasibility” using the updated feasibility model 25 .
- "feasibility 2" after the table (search table #2 for the third time) arranged in the upper right row in FIG. 10 corresponds to the value of feasibility derived using the feasibility model 25 after the second update. .
- the material searching and manufacturing system 1 performs processing based on the values of "priority1" and "feasibility2" in the third processing.
- the material searching and manufacturing system 1 updates the values of "priority” and "feasibility”, and uses the updated values to perform the fourth to sixth processes. is detailed.
- the material searching and manufacturing system 1 repeats the above-described processing until a molecule that satisfies desired physical property values is synthesized.
- the material searching and manufacturing system 1 designates materials in the form of a list containing at least one item, and selects materials in order of priority.
- the data structure of the search table #2 shown in FIG. 10 indicates an ID for identifying a material, priority information indicating a priority based on the predicted physical property value of the material, and feasibility of synthesizing the material. It includes feasibility information and composition success/failure information indicating success or failure of composition processing.
- the processing unit 202 of the control device 20 searches for a group of materials whose synthesis success/failure information indicates that they have not been processed, in descending order of priority based on the priority information. Perform a process to select materials that meet the criteria.
- FIG. 11 and 12 are diagrams showing detailed examples of processing using the physical property prediction model 24.
- FIG. 11 shows the details of the first to third processes in FIG. 9, and
- FIG. 12 shows the details of the fourth to sixth processes in FIG. Note that detailed description of the same points as in FIG. 9 will be omitted.
- uccess means success of synthesis and success of measurement (“YES” in S7 of FIG. 2)
- “failure” means failure of synthesis (see FIG. 2 (“NO” in S5 of FIG. 2) or measurement failure (“NO” in S7 of FIG. 2).
- the six tables shown in FIGS. 11 and 12 correspond to the search table #1 in FIG. It is different from the search table #1 in that it has items of "pass/fail” and "pass/fail of measurement”.
- the items “synthesis pass/fail” and “measurement pass/fail” in the candidate material table #1 store results of actual processing. Therefore, in the search table #1, the items “synthesis pass/fail” and “measurement pass/fail” are "n/a” (value undecided) for molecules that have not been processed.
- the material searching and manufacturing system 1 updates the items of "synthesis pass/fail” and “measurement pass/fail” of the candidate material table #1 according to the processing results for the processing target and lower molecules.
- the material searching and manufacturing system 1 uses the initial data #0 (data of the physical property database 22 before updating) to train the initial model #0 (physical property prediction model 24) by a learning algorithm.
- the material searching and manufacturing system 1 uses the initial model #0 to evaluate (derive) the priority of each molecule, and performs processing using the priority value.
- various parameters can be used in the evaluation (derivation) of priority. value is undecided) can be prioritized. By doing so, it is possible to increase the possibility of searching for new materials in consideration of feasibility.
- the material search and manufacturing system 1 performs processes such as synthesis on the molecule identified by the molecule ID "m100275" with the highest priority. Since the material searching and manufacturing system 1 failed to synthesize the molecule identified by the molecule ID "m100275", it changes the success/failure of synthesis of the molecule identified by the molecule ID "m100275" to "0". In addition, since the material searching and manufacturing system 1 failed to synthesize the molecules with the 2nd and 3rd priorities even in the second and third processes, it changes the success/failure of synthesizing the molecules with the 2nd and 3rd priorities to "0". .
- the material searching and manufacturing system 1 performs the synthesis process on the molecule identified by the molecule ID "m059059" with the fourth priority. It is assumed that the material searching and manufacturing system 1 has successfully synthesized and measured the molecule identified by the molecule ID "m059059". In this case, the material searching and manufacturing system 1 changes the synthesis pass/fail and measurement pass/fail of the molecule identified by the molecule ID "m059059" to "1" in the candidate material table.
- the material searching and manufacturing system 1 updates the physical property value information of the molecule identified by the molecule ID "m059059" in the physical property database 22. Further, the material searching and manufacturing system 1 re-learns (updates) the model #0 (physical property prediction model 24) using the data #1 (data of the physical property database 22 after updating). Then, the material searching and manufacturing system 1 updates the value of priority using the updated physical property prediction model 24 .
- the material searching and manufacturing system 1 performs the fifth and sixth processes in order from the first in order of priority based on the value of "priority1". For example, the material searching and manufacturing system 1 repeats the above-described processing until a molecule that satisfies desired physical property values is synthesized.
- FIG. 13 to 15 are diagrams showing detailed examples of processing using the physical property prediction model 24 and the feasibility model 25.
- FIGS. 13 to 15 show an example of a manufacturing method for manufacturing the physical property prediction model 24 and the feasibility model 25, which are machine learning models, based on the synthetic processing results.
- FIG. 13 shows details of the first and second processes in FIG. 10
- FIG. 14 shows details of the third and fourth processes in FIG. 10
- FIG. 15 shows the fifth process in FIG. Show the details of ⁇ 6 times. 10 to 12 will not be described in detail.
- uccess means success of synthesis and success of measurement (“YES” in S7 of FIG. 2)
- failure means failure of synthesis (see FIG. 2 (“NO” in S5 of FIG. 2) or measurement failure (“NO” in S7 of FIG. 2).
- the six tables shown in FIGS. 13 to 15 correspond to the search table #2 in FIG. ” and “Measurement pass/fail” items.
- the items “synthesis pass/fail” and “measurement pass/fail” in the candidate material table #2 store the results of the actual processing.
- the material searching and manufacturing system 1 uses the initial data (data of the physical property database 22 before updating) to train and update the physical property prediction model #0 (physical property prediction model 24) with a learning algorithm.
- the material searching and manufacturing system 1 evaluates (derives) the priority of each molecule using the updated physical property prediction model #0, and performs processing using the priority value.
- the material searching and manufacturing system 1 trains a feasibility model #0 (feasibility model 25) by a learning algorithm using initial data (data of the feasibility database 23 before updating). and update.
- the material searching and manufacturing system 1 evaluates (derives) the feasibility of each molecule using the updated feasibility model #0, and performs processing using the feasibility value.
- various parameters can be used in the evaluation (derivation) of feasibility, but priority is given to those with "n/a” (value undecided) for both "synthesis pass/fail” and “measurement pass/fail” items. You can also make it By doing so, it is possible to increase the possibility of searching for new materials in consideration of feasibility.
- n/a value undecided
- Material Exploration and Manufacturing System 1 does not process molecules with 1st to 3rd priority because the "feasibility" value is less than the threshold (0.5), and skips the process. Then, the material searching and manufacturing system 1 synthesizes, etc., the molecule identified by the molecule ID "m059059” whose priority is the fourth in the first process and whose "feasibility” value is equal to or greater than the threshold value (0.5). process. In FIG. 10, it is assumed that the molecule identified by the molecule ID "m059059” has been successfully synthesized and measured. In this case, the material searching and manufacturing system 1 changes the pass/fail of synthesis and pass/fail of measurement of the molecule identified by the molecule ID "m059059” to "1".
- the material searching and manufacturing system 1 updates the physical property value information of the molecule identified by the molecule ID "m059059" in the physical property database 22. Further, the material searching and manufacturing system 1 relearns (updates) the physical property prediction model #1 (the physical property prediction model 24) using the updated data of the physical property database 22. FIG. Then, the material searching and manufacturing system 1 updates the value of priority using the updated physical property prediction model 24 .
- the material searching and manufacturing system 1 adds (or updates) the information on the molecule identified by the molecule ID "m059059" to the feasibility database 23 as the information on the successfully synthesized molecule, thereby making the feasibility database 23 Update.
- the material searching and manufacturing system 1 uses the updated feasibility database 23 to re-learn (update) the feasibility model #1 (feasibility model 25). Then, the material searching and manufacturing system 1 updates the value of "feasibility" using the updated feasibility model #1.
- the material searching and manufacturing system 1 uses the molecule ID "m100686" which has the highest priority in the updated search table #2 and whose "feasibility" value is equal to or greater than the threshold value (0.5). Synthesis processing is performed on the identified molecule. In FIG. 10, it is assumed that the molecule identified by the molecule ID "m100686” has failed in synthesis. Therefore, the material searching and manufacturing system 1 changes the synthesis success/failure of the molecule identified by the molecule ID "m100686" to "0".
- the material searching and manufacturing system 1 updates the data. For example, the material searching and manufacturing system 1 adds (or updates) the information of the molecule identified by the molecule ID "m100686" to the feasibility database 23 as the information of the molecule whose synthesis failed, thereby making the feasibility database 23 Update again. As shown in FIG. 14, the material searching and manufacturing system 1 uses the updated feasibility database 23 to re-learn (update) the feasibility model #2 (feasibility model 25). Then, the material searching and manufacturing system 1 updates the value of "feasibility" using the updated feasibility model #2.
- the material searching and manufacturing system 1 performs processing based on the values of "priority1" and "feasibility2" in the third processing. As shown in FIGS. 14 and 15, the material searching and manufacturing system 1 updates the values of "priority” and "feasibility", and uses the updated values to perform the fourth to sixth processes. Since the process is the same as the process described above, detailed description will be given in detail, but the material searching and manufacturing system 1 repeats the process described above until a molecule that satisfies desired physical property values is synthesized. Specifically, in the sixth process, a molecule that satisfies desired physical property values is synthesized, and an example in which the target values are achieved in S8 of FIG. 2 is shown.
- the material searching and manufacturing system 1 changes the synthesis success/failure and measurement success/failure of the molecule identified by the molecule ID “m084127” to “1”, and furthermore, in the physical property database 22, the physical property value information of the molecule ID “m084127” is changed to Along with the update, the data of the molecule ID "m084127” is added to the feasibility database 23, and the pass/fail of synthesis and pass/fail of measurement are changed to "1". In this case, the process ends assuming that the destination has been reached.
- a feasibility model 25 which is a machine learning model, is manufactured. In the manufacturing method shown in FIGS.
- the control device 20 of the material searching and manufacturing system 1 selects materials according to the feasibility information output by the feasibility model 25.
- FIG. the control device 20 instructs the material manufacturing apparatus 10 to synthesize the selected materials, and based on the result of the synthesizing process from the material manufacturing apparatus 10, the feasibility database 23 (feasibility database) update the information in Then, the control device 20 updates the feasibility model 25 based on the feasibility information of the feasibility database 23 that has been updated.
- the manufacturing method shown in FIGS. 13-15 produces the updated feasibility model 25 .
- FIG. FIG.30 and FIG.31 is a figure which shows an example of the material manufacture result regarding a search.
- a material search utilizing the feasibility model 25 is performed.
- a material search was performed with the goal of optimizing physical property values, with 133,886 molecules included in the QM9 data set, which is a data set of organic molecular structural physical properties, set as the range of the material search space.
- the material manufacturing device 10 is realized by software (software material manufacturing device).
- the material manufacturing device 10 receives one molecular structure in the QM9 data set as an input from the control device 20, and outputs only when synthesis/measurement succeeds/failure (True, False) as output #1.
- the physical property values which are the measurement results, are returned to the controller 20 .
- the success or failure of the synthesis/measurement was determined by setting a threshold value for LogP (calculated value), which is a function that indicates the dissolution behavior with respect to the solvent. This corresponds to reproducing the material production fact that synthesis and measurement succeed/fail depending on the dissolution behavior with respect to the solvent, by simple calculation values.
- the physical property values as the measurement results, the physical property values calculated by first-principles calculation of a single molecule included in the QM9 data set were used.
- the control device 20 was also implemented as software (see 2 in Table 1).
- the search algorithm used by the search agent 26 includes two types of comparative examples (random search, search by the physical property prediction model 24) and an algorithm corresponding to the synthetic material selection method of the present disclosure (the physical property prediction model 24 and the executable Search by sex model 25) and each were implemented, and the performance was compared.
- Comparative Example #1 corresponds to random search
- comparative example #2 corresponds to search by physical property prediction model 24 .
- Procedure #1-1 Randomly select one molecule from unevaluated molecules contained in the material space, and synthesize and measure it. Mark the molecule as evaluated.
- Procedure #1-2 Synthesis/measurement (a): If synthesis/measurement fails: Repeat from procedure 1. (b): When the synthesis/measurement is successful: Store the molecular structure and measurement results (physical property values) in the physical property database 22 .
- Procedure #1-3 Completed when the maximum (minimum) measured value in the physical property database 22 reaches the target value. If not, repeat from step #1-1.
- Comparative example #2 search by physical property prediction model
- an algorithm called Bayesian optimization using Gaussian Process regression capable of calculating the prediction mean value ( ⁇ ) and prediction variance ( ⁇ ) was implemented as a machine learning model.
- Molecular descriptors eg, RdKit descriptors
- the search procedure is as shown in procedure #2 below.
- Step #2-1 Construct a physical property prediction model 24 (initial model) that gives equal outputs for all inputs.
- Step #2-3 Prioritize the unevaluated molecules contained in the material space based on the acquisition function Fa.
- Step #2-4 Synthesize and measure the molecule with the highest priority. Mark this molecule as evaluated.
- Procedure #2-5 Synthesis/measurement (a): If synthesis/measurement fails: Repeat from procedure #2-4. (b): When the synthesis/measurement is successful: Store the molecular structure and measurement results (physical property values) in the physical property database 22 and update the physical property prediction model 24 .
- Procedure #2-6 Completed when the maximum (minimum) value of the measured values in the physical property database 22 reaches the target value. If not, repeat from step #2-2.
- Example #1 and Example #2 corresponding to the synthetic material selection method of the present disclosure will be described. Note that points similar to those described above will be described as appropriate.
- Example #1 and Example #2 Search by physical property prediction model and feasibility model
- Gaussian Process regression was used as the machine learning model for the property prediction model 24 and Gaussian Process classification was used as the machine learning model for the feasibility model 25 .
- Molecular descriptors eg, RdKit descriptors
- the search procedure is as shown in procedure #3 below.
- the initial model of feasibility model 25 is model built using evaluated data (successfully synthesized data).
- Procedure #3-1 A feasibility model 25 (initial model) is constructed using initial data (number of data: 100 or 10).
- Step #3-2 Construct a physical property prediction model 24 (initial model) that gives equal outputs for all inputs.
- Step #3-4 Prioritize the unevaluated molecules contained in the material space based on the acquisition function Fa.
- Procedure #3-5 The feasibility model 25 evaluates the feasibility (prediction of success or failure of synthesis/measurement) of the unevaluated molecules contained in the material space.
- Step #3-6 Synthesize and measure molecules that have high feasibility (probability of 0.5 or more, etc.) and have the highest priority. Mark this molecule as evaluated.
- Procedure #3-7 Synthesis/measurement (a): If synthesis/measurement fails: Update feasibility database 23 and feasibility model 25, and repeat from procedure #3-5. (b): When the synthesis/measurement is successful: update the feasibility database 23 and the feasibility model 25, and update the physical property database 22 and the physical property prediction model 24 from the molecular structure and measurement results (physical property values).
- Procedure #3-8 Completed when the maximum value (minimum value) of the measured values held in the physical property database 22 reaches the target value. If not, repeat from step #3-2.
- the evaluation index (see 6 in Table 1) is the number of material manufacturing trials to reach the target value, 1,000 for Comparative Example #1, Comparative Example #2 and Example #1 and Example For #2, 100 trials were performed and average values were compared (see Table 1 for detailed settings).
- Example #1 and Example #2 the initial data used to construct the initial model of the feasibility model 25 was given (number of data: 100 or 10). This corresponds to a case in which, as prior knowledge of the material manufacturing apparatus 10, the result of whether or not the synthesis/measurement is successful for 100 or 10 molecular structures included in the initial data is known.
- the initial model of the feasibility model 25 is constructed using teacher data for 100 molecular structures.
- the initial model of the feasibility model 25 is constructed using teacher data for 10 molecular structures.
- FIGS. 30 and 31 show the evaluation results of molecular structure searches aimed at optimizing (maximizing) Heat Capacity.
- the target value (45.0 kcal/mol or more) is the condition for comparison in the achievement of the target value in step S8 of FIG.
- the search by the physical property prediction model 24 (comparison For example #2)
- efficiency can be improved by a maximum of 13 times or more (1/13 of the time and cost). Therefore, in the synthetic material selection method and material manufacturing method of the present disclosure, material synthesis can be performed efficiently based on the feasibility of synthesis.
- FIG. 16 to 21 are diagrams showing an example of the feasibility database 23.
- the feasibility database 23 may be the database DB4 in FIG.
- the database DB4 includes items such as "molecule ID” and "synthesis & measurement pass/fail".
- “Molecule ID” indicates identification information for identifying a molecule.
- “Synthesis & measurement pass/fail” indicates whether or not synthesis and measurement of the molecule could be executed. For example, “synthesis & measurement pass/fail” is “1” when both synthesis and measurement of a molecule can be executed, and "0” otherwise.
- the physical property prediction model 24 is also updated. corresponds to "success" in the example of
- the feasibility database 23 may be the database DB5 in FIG.
- the database DB5 includes items such as "molecule ID” and "synthesis yield".
- “Molecule ID” indicates identification information for identifying a molecule.
- “Synthetic yield” indicates the yield in the synthesis of the molecule.
- the feasibility database 23 may be the database DB6 in FIG.
- the database DB6 includes items such as "molecule ID”, “synthesis pass/fail”, “preparation pass/fail”, and “absorption spectrum measurement pass/fail”.
- “Molecule ID” indicates identification information for identifying a molecule.
- “Synthesis pass/fail” indicates whether or not the synthesis of the molecule could be executed.
- “Fragment of fractionation” indicates whether fractionation of molecules could be executed.
- “Absorption spectrum measurement pass/fail” indicates whether or not the absorption spectrum of the molecule could be measured.
- each of "synthesis acceptance/rejection”, “preparation acceptance/rejection” and “absorption spectrum measurement acceptance/rejection” is “1” if the process can be executed, and is “0” otherwise.
- “absorption spectrum measurement pass/fail” is an example of the item “measurement pass/fail” shown in the embodiments of FIGS. 13 to 15 .
- the feasibility database 23 may be the database DB7 in FIG.
- the database DB7 includes items such as "molecule ID”, "solubility in solvent A”, “solubility in solvent B”, “physical property A measurement”, and “physical property B measurement”.
- “Molecule ID” indicates identification information for identifying a molecule.
- “Synthesis pass/fail” indicates whether or not the synthesis of the molecule could be executed.
- “Solvent A solubility” indicates a value for the solubility of the molecule in solvent A.
- solvent B solubility indicates a value for the solubility of the molecule in solvent B;
- Physical property A measurement indicates whether or not the physical property A can be measured.
- “Physical property B measurement” indicates whether or not physical property B could be measured. For example, each of “synthesis pass/fail”, “physical property A measurement”, and “physical property B measurement” is “1" when the processing can be executed, and is "0” otherwise.
- the feasibility database 23 may be the database DB8 in FIG.
- the database DB8 includes items such as "molecule ID”, “synthesis path ID”, “synthesis pass/fail”, “fractionation pass/fail”, and “absorption spectrum measurement pass/fail”.
- “Molecule ID” indicates identification information for identifying a molecule.
- “Synthesis path ID” indicates identification information for identifying a synthesis path (route) of a molecule.
- Synthesis pass/fail indicates whether or not the synthesis of the molecule could be executed.
- “Fragment of fractionation” indicates whether fractionation of molecules could be executed.
- “Absorption spectrum measurement pass/fail” indicates whether or not the absorption spectrum of the molecule could be measured.
- the feasibility database 23 may be the database DB9 in FIG.
- the database DB9 includes items such as "synthetic path ID", "reaction step 1", and "reaction step 2".
- “Synthesis path ID” indicates identification information for identifying a synthesis path (route) of a molecule.
- “Reaction step 1” indicates the first step in the synthesis path identified by the synthesis path ID.
- “Reaction Step 2” indicates the second step in the synthesis pass identified by the synthesis pass ID. Although only “reaction step 1" and “reaction step 2" are illustrated in FIG. have.
- the material searching and manufacturing system 1 can execute arbitrary processing using the various types of information described above. This point will be described below as another embodiment different from the above-described embodiment.
- the material searching and manufacturing system 1 calculates a priority based on information on physical properties (characteristics) and/or feasibility, and creates a synthetic material (molecule) list (including at least one) according to the priority. Each item in the list contains at least a molecule ID. Note that the list may include ⁇ molecule ID, synthetic path ID ⁇ pairs.
- the synthesizing device 11 sends a material synthesis list to the synthesizing device 11 via the control interface 21, and instruct synthesis according to the list.
- the synthesizing device 11 reads ⁇ molecule ID (, synthetic pass ID) ⁇ according to the priority of the list.
- the synthesizing device 11 uses the synthesis pass ID as a key to acquire reaction steps from the synthesis pass ID table, and performs synthesis processing.
- the synthesizer 11 If the conditions specified in the reaction step are satisfied, the synthesis process is performed. (4-1) When the synthesizer 11 receives the notification of "synthesis not possible", the synthesizer 11 notifies the control interface 21 of synthesis failure for the first molecule ID (synthesis failure in FIG. 2). , go to (6). (4-2) Otherwise, the synthesizer 11 performs the reaction step 1 process. As a result of processing, either "normal termination” or "abnormal termination” is notified. (4-3) In the case of "abnormal termination", the synthesizer 11 notifies the control interface 21 of synthesis failure for the first molecule ID (synthesis failure in FIG.
- Synthesizer 11 confirms whether there is a next reaction step in the case of "normal completion", and if there is a next reaction step, acquires the next reaction step, and from (4) repeat. If there is no next reaction step, proceed to (5).
- the synthesizing device 11 notifies the control interface 21 of success or failure of synthesizing.
- the "success/failure notification” differs depending on the notification received during the synthesis process: (6-1) In the case of synthesis failure, abnormal termination, or synthesis failure, transmit "synthesis failure”. (6-2) In the case of success in synthesis, transmit "success in synthesis”.
- the synthesizer 11 If there is a material in the list for which synthesis has not yet been attempted, the synthesizer 11 reads out the next ⁇ molecule ID (, synthesis pass ID) ⁇ and repeats from (3). (8) In parallel with the processing of the synthesizing device 11, the search agent 26 determines that synthesizing is impossible in the case of "synthesis failure", and builds or updates the "feasibility model".
- FIG. 22 shows the detailed flow of the processes (1) to (8) described above.
- FIG. 22 is a sequence diagram showing an example of processing procedures in the material searching and manufacturing system. It should be noted that the step numbers shown in FIG. 22 are given for the sake of convenience in explaining the processing by each processing subject, and do not prescribe the order of the processing.
- the search agent 26 transmits the selected material synthesis list to the synthesis device 11 via the control interface 21 (step S101).
- the search agent 26 instructs the synthesizer 11 to synthesize the materials specified in the list (step S102).
- the search agent 26 waits for result information of one material synthesis in the list from the synthesizing device 11 (step S103).
- the search agent 26 processes the result in accordance with the judgment in step S5 of FIG. 2 (step S104).
- step S105: No If the search agent 26 has not finished synthesizing all the molecule IDs shown in the list (step S105: No), it returns to step S103 and repeats the process. Further, the search agent 26 ends the process when the synthesis process has been completed for all the molecule IDs shown in the list (step S105: Yes).
- control interface 21 When the control interface 21 receives “unable to combine”, “abnormal termination”, or “failure to combine” from the combining device 11 , it transmits “failure to combine” to the search agent 26 and receives “successful combining” from the combining device 11 . If so, it transmits "synthesis failure" to the search agent 26 (step S111).
- the synthesizer 11 receives the material synthesis list ⁇ molecule ID,... ⁇ (step S121).
- the synthesizing device 11 reads the target ⁇ molecule ID (, synthesis path ID) ⁇ according to the priority of the list (step S122). If the conditions for the reaction step specified by the synthesis path ID are not satisfied (step S123: No), the synthesis device 11 transmits "cannot be synthesized" to the control interface 21 and selects the next synthesis target (step S120). ), return to step S122 and repeat the process.
- step S123: Yes When the conditions for the reaction step specified by the synthesis pass ID are satisfied (step S123: Yes), the synthesizer 11 synthesizes the material for the molecule ID (step S124). When step S124 ends abnormally (step S125: Yes), the synthesizing device 11 transmits "abnormal end" to the control interface 21, selects the next synthesis target (step S120), and returns to step S122 for processing. repeat.
- step S124 did not end abnormally (step S125: No)
- step S126 determines whether there is a next reaction step (step S126). If there is a next reaction step (step S126: Yes), the synthesizer 11 selects the next reaction step (step S127), returns to step S123, and repeats the process.
- step S126 If there is no next reaction step (step S126: No), the synthesizing device 11 instructs the analyzing device 12 to perform analysis processing.
- the synthesizing device 11 notifies the analysis result (step S128).
- the synthesizing device 11 notifies the search agent 26 via the control interface 21 of information corresponding to "synthesis success" in case of success and "synthesis failure” in case of failure.
- step S129: Yes If the list still contains the next molecule ID to be synthesized (step S129: Yes), the synthesizer 11 returns to step S120 and repeats the process. If there is no molecule ID to be synthesized next in the list (step S129: No), the synthesizer 11 ends the process.
- the analysis device 12 receives the synthesized material from the synthesis device 11 and performs analysis processing (step S131). The analysis device 12 then transmits the analysis result to the synthesis device 11 .
- the material searching and manufacturing system 1 executes the following processes (11) to (16).
- This measurement process corresponds to the "measurement" process (step S6) when the search agent performs the process shown in FIG. It is an example of a series of processing performed in .
- the search agent 26 receives the successfully synthesized molecule ID. (12) Specifying the ID of the successfully synthesized molecule, designating the physical properties to be measured, and instructing the synthesizer 11 via the control interface 21 to measure the successfully synthesized material. (13) Synthesizing device 11 that has received the instruction automatically controls to provide measuring device 13 with the synthetic material specified by the instructed molecule ID. Instruct to start measuring physical properties.
- the measuring device 13 notifies the synthesizing device 11 of "measurement not possible” when the measurement cannot be performed, and of "measurement completed” and the measurement result when the measurement is completed normally. 11 transmits measurement results via the control interface 21 . (15) If the measurement result is "impossible to measure", the search agent 26 determines that synthesis cannot be performed, and builds or updates the "feasibility model".
- the search agent 26 further determines whether the target value has been achieved, and if the target value has been achieved, determines that the synthesis has succeeded, and constructs or updates the "physical property prediction model". I do. Also, if the target value is not achieved, it is determined that synthesis cannot be performed, and the "feasibility model" is constructed or updated.
- FIG. 23 is a sequence diagram showing an example of processing procedures in the material searching and manufacturing system. It should be noted that the step numbers shown in FIG. 23 are given for the sake of convenience in explaining the processing by each processing entity, and do not prescribe the order of the processing.
- the search agent 26 transmits the successfully synthesized molecule ID to the synthesizer 11 (step S201).
- the search agent 26 designates the physical properties to be measured to the synthesizer 11 and instructs the measurement of the material (step S202).
- the search agent 26 waits for result information of one material synthesis in the list from the synthesizing device 11 (step S203).
- the search agent 26 processes the result in accordance with the judgment of step S7 in FIG. 2 (step S204), and terminates the process.
- control interface 21 When the control interface 21 receives "measurement impossible”, it transmits "measurement failure" to the search agent 26. When it receives "measurement end”, it transmits "measurement success” to the search agent 26 (step S211). .
- the synthesizer 11 receives the molecule ID (step S221).
- the synthesizer 11 provides the synthesized material designated by the molecule ID to the measuring device 13 and instructs measurement of the designated physical property (step S222).
- the synthesizing device 11 notifies the measurement result (step S223).
- the synthesizing device 11 notifies the search agent 26 via the control interface 21 of information corresponding to "measurement completed" in case of success and "measurement not possible” in case of failure.
- the measuring device 13 performs measurement processing of the synthesized material for the physical property specified by the synthesizing device 11 (step S231). The measuring device 13 then transmits the measurement result to the synthesizing device 11 .
- the synthesis device 11 receives information/data/instructions via the control interface 21, and the synthesis device 11 receives the information/data/instructions.
- Information, data, and instructions are transmitted to the measuring device 13, and the synthesizing device 11 receives the result information from the analyzing device 12 or the measuring device 13.
- the measurement device 13 may directly exchange data with the control interface 21 .
- the search agent 26 (or equivalent computer software or hardware) may be allowed to directly control the analysis device 12 or the measurement device 13 via the control interface 21.
- a computer such as the information processing apparatus 100 shown in FIG. 24 may have the function of the search agent 26 .
- FIG. 24 is a diagram showing an example of the configuration of a material searching and manufacturing system.
- the control interface 21 may be a computer device (control device 200 or the like in FIG. 24) having a communication interface, or may be a wired/wireless communication device. Control device 200 functions as control device 20 .
- the information processing apparatus 100 is connected to other configurations via a network N1 such as the Internet (cloud) or a network N2 such as a private cloud including virtualization technology and VPN (Virtual Private Network) technology. connect.
- a network N1 such as the Internet (cloud) or a network N2 such as a private cloud including virtualization technology and VPN (Virtual Private Network) technology.
- the information processing device 100 communicates with the material manufacturing device 10, database DB, etc. via the control device 200 .
- the database DB may operate on a computer connected to the synthesizing device 11 via a network, on a computer on a cloud, or on a device integrated with the synthesizing device 11 .
- the database DB may include the same database as the physical property database 22 (physical property DB in FIG. 1) or the performance database 23 (performance DB in FIG. 1).
- the database DB may be on separate storages, like the physical property database 22 (physical property DB in FIG. 1) and the performance database 23 (performance DB in FIG. 1), or may be in one storage. It may be integrated as one database DB, or may be on a virtual storage device.
- control interface 21 may have a function of accessing the database, and may perform the processing related to the material synthesis list that was performed by the synthesizing device 11 in the synthesizing phase.
- the synthesizing device 11, the analyzing device 12, and the measuring device 13 may be a system with the same control system in one housing, or may be independent in separate housings.
- FIG. 25 is a diagram illustrating a configuration example of an information processing apparatus according to the present disclosure;
- the information processing device 100 functions as the control device 20 .
- the information processing device 100 may be integrated with the control device 200 .
- the information processing device 100 has a communication section 110, a storage section 120, and a control section .
- the information processing apparatus 100 includes an input unit (for example, a keyboard, a mouse, etc.) that receives various operations from an administrator of the information processing apparatus 100, and a display unit (for example, a liquid crystal display, etc.) for displaying various information. may have.
- the communication unit 110 is realized by, for example, a NIC (Network Interface Card) or the like.
- the communication unit 110 is connected to a predetermined network (not shown) by wire or wirelessly, and transmits and receives information to and from each component of the material searching and manufacturing system 1 such as the material manufacturing apparatus 10 and the control device 20 .
- the communication unit 110 functions as a control interface 21 that controls the material manufacturing device 10 by transmitting information to the material manufacturing device 10 .
- the communication unit 110 controls synthesis processing by transmitting and receiving information to and from the material manufacturing apparatus 10 .
- the communication unit 110 instructs the material manufacturing apparatus 10 to perform synthesis processing.
- the communication unit 110 instructs the material manufacturing apparatus 10 to perform the synthesis process by transmitting information indicating materials to be subjected to the synthesis process.
- the communication unit 110 receives the synthetic processing result from the material manufacturing apparatus 10 .
- the communication unit 110 receives the physical property measurement processing result from (the measuring device 13 of) the material manufacturing apparatus 10 .
- the storage unit 120 is implemented by, for example, a semiconductor memory device such as RAM (Random Access Memory) or flash memory, or a storage device such as a hard disk or optical disk.
- the storage unit 120 includes a physical property database 22 that holds physical property data, a feasibility database 23 that holds whether synthesis and/or measurement can be executed, a physical property prediction model 24 that predicts physical properties, and whether synthesis and/or measurement can be executed. It stores various information such as a feasibility model 25 that predicts what will happen.
- the control unit 130 uses, for example, a CPU (Central Processing Unit) or an MPU (Micro Processing Unit) to operate a program stored inside the information processing apparatus 100 (for example, an information processing program according to the present disclosure, etc.) in RAM, etc. It is realized by being executed as a region. Also, the control unit 130 is implemented by an integrated circuit such as an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array).
- ASIC Application Specific Integrated Circuit
- FPGA Field Programmable Gate Array
- the control unit 130 uses the information received from the external device by the communication unit 110 to execute each process. Also, the control unit 130 executes each process using the information stored by the storage unit 120 . Control unit 130 controls material manufacturing apparatus 10 by transmitting information to material manufacturing apparatus 10 via communication unit 110 .
- the control unit 130 executes processing corresponding to the functions of the search agent 26 .
- the control unit 130 executes the processing related to the synthetic material selection described above.
- the control unit 130 executes the processing related to the production of the material described above.
- the control unit 130 instructs the material manufacturing apparatus 10 to synthesize materials, thereby causing the material manufacturing apparatus 10 to perform processing for manufacturing materials.
- the control unit 130 selects a material from among the materials stored in the storage unit 120 using the material physical property information and the feasibility information.
- the control unit 130 instructs the material manufacturing apparatus 10 to synthesize the selected materials.
- the control unit 130 controls the material synthesizing process by the material manufacturing apparatus 10 by transmitting list-format information to the material manufacturing apparatus 10 in accordance with the order of priority.
- the control unit 130 controls the material synthesizing process by the material manufacturing apparatus 10 by transmitting the search table #2 to the material manufacturing apparatus 10 .
- the control unit 130 designates materials to be synthesized by transmitting list-format information including designation of materials to be synthesized (materials to be synthesized) to the material manufacturing apparatus 10 .
- control unit 130 designates a material to be synthesized by transmitting a search table #2 in which a flag is attached to the material to be synthesized to the material manufacturing apparatus 10 .
- Control unit 130 updates the feasibility information stored in storage unit 120 based on the synthesis processing result from material manufacturing apparatus 10 .
- the control unit 130 updates the information in the search table #2 based on the synthesizing processing result from the material manufacturing apparatus 10 .
- the control unit 130 selects a material from two or more synthesis processing target materials stored in the storage unit 120 .
- the control unit 130 selects materials so as not to include materials that have been successfully synthesized in the past from materials to be synthesized.
- the control unit 130 refers to information stored in the storage unit 120, excludes materials that have been successfully synthesized in the past, and selects materials. This allows the control unit 130 to select a new material.
- the control unit 130 selects the material according to the priority based on the first predicted value of the material based on the material physical property information.
- the control unit 130 performs priority material selection based on the second predicted value of the material based on the feasibility information.
- the control unit 130 selects materials whose second predicted values are equal to or greater than a predetermined threshold.
- control unit 130 updates the feasibility information stored in the storage unit 120 for the successfully synthesized materials.
- the control unit 130 updates the feasibility model 25 based on the updated feasibility information.
- the control unit 130 designates the identifier of the successfully synthesized material and the physical properties to be measured, and instructs the physical property measurement process.
- the control unit 130 updates the material physical property information.
- the control unit 130 updates the material property information of the successfully synthesized material in the storage unit 120 based on the property measurement processing result from the material manufacturing apparatus 10 .
- the control unit 130 updates the material physical property information for the material that satisfies the target value stored in the storage unit 120 .
- the control unit 130 updates the physical property prediction model 24 based on the updated material physical property information.
- the control unit 130 controls the instructed physical property measurement processing for the successfully synthesized material identified by the identifier, and provides the identifier and the measured physical property value.
- the control unit 130 causes another device to transmit the physical property value via the communication unit 110 .
- the control unit 130 transmits information indicating the identifier of the successfully synthesized material and the measured physical property value to another device via the communication unit 110 .
- the control unit 130 performs material selection using data having a synthetic material selection data structure (also referred to as “synthetic material selection data”). For example, the control unit 130 selects materials using data (synthetic material selection data) such as the search table #2 in FIG. 10 or the candidate material table #2 in FIGS.
- the control unit 130 includes an ID (for example, a molecule ID) that identifies a material, priority information (for example, a priority value) indicating a priority based on the predicted physical property value of the material, and Materials are selected using synthetic material selection data including feasibility information (for example, feasibility value) indicating feasibility and synthesis success/failure information (for example, synthesis success/failure value) indicating success or failure of synthesis processing.
- the control unit 130 uses the synthesis material selection data, the control unit 130 searches for a group of materials in which the synthesis success/failure information indicates unprocessed materials in descending order of priority based on the priority information, and performs synthesis processing indicated by the synthesis success/failure information.
- control unit 130 searches for a group of molecules whose synthesis pass/fail value is unknown (n/a) in order from the highest priority value, and selects materials whose feasibility value is greater than or equal to a predetermined threshold. do.
- FIG. 26 to 28 are diagrams showing an example of the configuration of a material manufacturing apparatus.
- material manufacturing apparatuses 10A, 10B, and 10C will be described according to the mode of the material moving mechanism. Further, regarding the material manufacturing apparatus 10, descriptions of the same points as those described above will be omitted as appropriate.
- a material manufacturing apparatus 10A shown in FIG. 26 has a material moving path 14 as a material moving mechanism.
- the synthesizing device 11 has a driving device 111 that drives the material moving path 14, a processor 112 that controls the driving device 111 and performs synthesizing processing, and a communication interface 113 for communicating with other devices.
- the analysis device 12 also has a drive device 121 that drives the material transfer path 14, a processor 122 that controls the drive device 121 and executes analysis processing, and a communication interface 123 for communicating with other devices.
- the measuring device 13 also has a driving device 131 that drives the material moving path 14, a processor 132 that controls the driving device 131 and performs measurement processing, and a communication interface 133 for communicating with other devices.
- FIG. 26 is a belt conveyor or the like provided over the synthesizing device 11 , the analyzing device 12 and the measuring device 13 .
- the material MT shown in FIG. 26 can move among the synthesizing device 11, the analyzing device 12 and the measuring device 13 according to the operation of the material moving path 14.
- FIG. 26 is a belt conveyor or the like provided over the synthesizing device 11 , the analyzing device 12 and the measuring device 13 .
- the material MT shown in FIG. 26 can move among the synthesizing device 11, the analyzing device 12 and the measuring device 13 according to the operation of the material moving path 14.
- the material manufacturing apparatus 10B shown in FIG. 27 has a robot arm 15 as a material moving mechanism.
- the synthesizing device 11 has a driving device 111 for driving the robot arm 15, a processor 112 for controlling the driving device 111 and performing synthesizing processing, and a communication interface 113 for communicating with other devices.
- the analysis device 12 also has a processor 122 for executing analysis processing and a communication interface 123 for communicating with other devices.
- the measuring device 13 also has a processor 132 for executing measurement processing and a communication interface 133 for communicating with other devices.
- the robot arm 15 shown in FIG. 27 is driven by the driving device 111 of the synthesizer 11 .
- the material MT shown in FIG. 27 can be moved among the synthesizing device 11 , the analyzing device 12 and the measuring device 13 according to the motion of the robot arm 15 .
- the material manufacturing apparatus 10C shown in FIG. 28 has a flow path with a syringe pump 16 and a tube 17 as a material moving mechanism.
- the synthesizing device 11 has a driving device 111 that drives the syringe pump 16, a processor 112 that controls the driving device 111 and performs synthesizing processing, and a communication interface 113 for communicating with other devices.
- the analysis device 12 also has a processor 122 for executing analysis processing and a communication interface 123 for communicating with other devices.
- the measuring device 13 also has a processor 132 for executing measurement processing and a communication interface 133 for communicating with other devices.
- the syringe pump 16 shown in FIG. 28 is driven by the driving device 111 of the synthesizing device 11 .
- the above is merely an example, and the material manufacturing apparatus 10 may have any form of material moving mechanism as long as the material can be moved between apparatuses.
- the material manufacturing apparatus 10 executes processing for manufacturing materials.
- the material manufacturing apparatus 10 executes a process of manufacturing materials according to an operation by a user of the material manufacturing apparatus 10 or the like.
- the material manufacturing apparatus 10 executes processing for manufacturing materials according to instructions from an external device.
- the material manufacturing device 10 executes processing for manufacturing materials according to instructions from the control device 20 .
- the material manufacturing apparatus 10A shown in FIG. 26 the material manufacturing apparatus 10B shown in FIG. 27, or the material manufacturing apparatus 10C shown in FIG. Targeting the moved material, a process of measuring the specified physical property value is performed.
- the measurement device 13 transmits the physical property measurement processing result generated by the measurement processing to another device.
- the material manufacturing device 10 transmits the physical property measurement processing result to other devices such as the control device 20, the information processing device 100, and the control device 200.
- the synthetic material selection method according to the present disclosure is based on material physical property information in the database (in the above example, information shown in database DB1, etc.; the same applies hereinafter) and feasibility information (in the above example, information shown in database DB2, etc.). hereinafter the same) is used to select the material to be synthesized, and the control device (in the above example, the material manufacturing device 10, the control device 20, the information processing device 100 or the control device 200, etc.; the same applies hereinafter), Instructions are given for synthesis processing of the selected materials, and feasibility information in the database is updated based on the synthesis processing results from the control device.
- material physical property information in the database in the above example, information shown in database DB1, etc.; the same applies hereinafter
- feasibility information in the above example, information shown in database DB2, etc.
- the synthetic material selection method executed by the material searching and manufacturing system 1 includes selecting materials to be synthesized using material property information and feasibility information in the database; directing the synthesis process of the materials and updating the feasibility information in the database based on the synthesis process results. This allows the synthetic material selection method to enable material synthesis based on synthetic feasibility.
- the database includes two or more synthetic processing target materials.
- the synthesis material selection method can select a material using a database containing two or more synthesis process target materials.
- the synthetic material selection method in material selection in the synthetic material selection method according to the present disclosure, selection is made so as not to include materials that have been successfully synthesized in the past. As a result, the synthetic material selection method can select materials so as not to include materials that have been successfully synthesized in the past.
- material selection in the synthetic material selection method according to the present disclosure is performed according to the priority based on the first predicted value of the material based on the material physical property information. This allows the synthetic material selection method to select materials according to priorities based on the first predicted value of the material.
- the first predicted value is input to the physical property prediction model (physical property prediction model 24 in the above example; the same shall apply hereinafter), and constitutes the material registered in the database. is output based on the material feature amount of each material.
- the synthetic material selection method can select the material according to the priority based on the first predicted value of the material predicted by the physical property prediction model.
- material selection in the synthetic material selection method according to the present disclosure is performed according to the priority based on the second predicted value of the material based on the feasibility information.
- the synthetic material selection method can select the material according to the second predicted value of the material based on the feasibility information.
- materials are selected from those whose second predicted values are equal to or greater than a predetermined threshold. This allows the synthetic material selection method to make an appropriate material selection.
- the second predicted value is input to the feasibility model 25, based on the material feature amount of each of the two or more materials that constitute the material registered in the database output.
- the synthetic material selection method can select the material according to the second predicted value of the material based on the feasibility information.
- the selected materials are selected in order of priority, and the selected materials are specified in a list format including at least one item, the items being material identifiers and Includes priority. This allows the synthetic material selection method to appropriately select materials.
- control device controls synthesis processing of materials provided in list form according to priority, provides synthesis processing results for each material, and the synthesis processing results are controlled. It is the identifier of the material of the synthesis process and the success/failure information of synthesis. This allows the synthetic material selection method to appropriately select materials.
- the control of the synthesis process instructs the synthesis device (the synthesis device 11 or the material manufacturing device 10 in the above example; the same applies hereinafter) to perform synthesis processing, and the synthesis device outputs the synthesis processing result including processing to receive This allows the synthesis material selection method to make available the results of the synthesis process.
- the feasibility information in the synthetic material selection method according to the present disclosure, if the synthesis success or failure information is successful, the feasibility information is updated for the successfully synthesized material in the database, and further updated. Based on the feasibility information, the feasibility model 25 is updated. This allows the synthetic material selection method to make appropriate material selections using the updated model.
- the controller based on the success or failure information of synthesis, the controller is instructed to perform physical property measurement processing by designating the identifier of the successfully synthesized material and the physical property to be measured, and controlling It includes a process of updating the material property information of the successfully synthesized material in the database based on the property measurement processing result from the device. This allows the synthetic material selection method to select an appropriate material using the updated material property information.
- the material physical property information when the physical property measurement result satisfies the target value, the material physical property information is updated in the database for the material that satisfies the target value. It includes processing to update the physical property prediction model based on the material physical property information. This allows the synthetic material selection method to select an appropriate material using the updated physical property prediction model.
- the control device controls the instructed physical property measurement processing for the successfully synthesized material specified by the identifier, and provides the identifier and the measured physical property value. Accordingly, the synthetic material selection method can provide information indicating appropriately measured physical property values and materials corresponding to the physical property values.
- the control of the physical property measurement process automatically moves the material synthesized by the synthesizing device to the measuring device (the measuring device 13 in the above example, the same applies hereinafter), and the instructed physical property It includes a process of instructing the measuring device to perform the measurement and receiving the physical property measurement processing result from the measuring device. This allows the synthetic material selection method to make available the results of the physical property measurement process.
- the attributes of the material are low molecular weight, dye, high molecular weight, fluorescent/identifying isotope labeling, self-organizing material/structure, biomaterial (sugar, peptide, polypeptide, amino acids, proteins, fatty compounds, DNA, etc.), organic thin films (evaporation, coating processes), inorganic materials (solid phase method, coprecipitation method, melt quenching method, sol-gel method, etc.), nanoparticles, metal complexes, inorganic thin films (ALD , sputtering, etc.), synthetic materials based on synthetic biological techniques (genetic recombination and material synthesis using bacteria), and functional materials with crystal structures, nanostructures, and microstructures. This allows the synthetic material selection method to appropriately select materials using information about the attributes of the materials.
- the material physical property information in the database is structured to include at least two items, and one of the at least two items is an attribute item of the material.
- the synthetic material selection method can appropriately select the material using the information regarding the attribute of the material physical property information including two or more items.
- the material manufacturing method performs material selection for selecting a material to be synthesized using material physical property information and feasibility information in the database, and instructs the control device to select the selected material.
- Materials are manufactured by a process of instructing a synthesis process and updating the feasibility information in the database based on the synthesis process results from the controller. This allows the synthetic material selection method to produce synthesized materials based on synthetic feasibility.
- the synthesis material selection data structure is an information processing device (in the above example, the control device 20, the information processing device 100, or the control device 200, etc.) that selects a material to be subjected to synthesis processing. ), which includes an ID that identifies the material, priority information that indicates the priority based on the predicted physical property value of the material, and feasibility that indicates the feasibility of synthesizing the material.
- an information processing device searches for a group of materials in which the synthesis success/failure information indicates that the synthesis success/failure information indicates that the material has not been processed in descending order of priority based on the priority information,
- the success or failure of the synthesizing process indicated by the synthesizing success/failure information is used in the process of selecting materials that satisfy a predetermined criterion. This allows the synthetic material selection method to enable material synthesis based on synthetic feasibility.
- the material searching and manufacturing system 1 avoids materials that are likely to fail in synthesis/measurement and reduces unnecessary synthesis/measurement, thereby dramatically improving efficiency in terms of time and cost. material search and manufacturing system.
- the material searching and manufacturing system 1 based on the information in the current database, a new material that has not been synthesized in the past is selected as a synthesis target, and the result of attempting synthesis with an actual synthesis apparatus is used.
- the data can be updated in real time using the results of measuring whether the material with the expected properties can be produced, The results are used to construct a feasibility model 25 and a physical property prediction model. Therefore, when selecting materials next time, avoiding materials that are likely to fail in synthesis and measurement and reducing unnecessary synthesis and measurement will dramatically improve efficiency in terms of time and cost.
- a material search and manufacturing system can be constructed.
- the material searching and manufacturing system 1 is a high-precision physical property prediction system that utilizes a physical property database that stores physical property values with matching material manufacturing conditions.
- the material searching and manufacturing system 1 is provided with a database storing physical property values, and in conjunction with the synthesizing device, the physical property prediction is performed so that the data can be updated based on the latest measurement data of the synthesized material. Models can be built and updated. Therefore, the synthesis material selection method has the effect of being able to predict the physical properties of materials that can be newly synthesized with higher accuracy than conventional methods.
- the material searching and manufacturing system 1 utilizes the feasibility database 23 that accumulates feasibility data (success or failure of synthesis and measurement) with matching material manufacturing conditions. It is a prediction system.
- a new material is selected based on the information in the database at the time of material selection, and the database is updated in real time using the result of trying to synthesize with an actual synthesizer.
- the synthesis is successful, it is possible to update the data in real time using the results of measuring whether the material with the expected properties was produced, and to use the results to improve feasibility.
- a model 25 and a physical property prediction model 24 are constructed. Therefore, it is possible to construct a material synthesis possibility prediction system and a measurement success probability prediction system by utilizing the feasibility database 23 that accumulates feasibility data (success or failure of synthesis and measurement) with matching material manufacturing conditions.
- the material search and manufacturing system 1 can be used for low molecular weight, dye, high molecular weight, fluorescent/identifying isotope labeling, self-organizing materials/structures, biomaterials (sugars, peptides, polypeptides, amino acids, proteins, fats, etc.).
- the material search and manufacturing system 1 is a structure search system that synthesizes, measures, and optimizes the properties of functional materials with crystal structures, nanostructures, and microstructures, and is capable of producing stable and metastable crystal structures and aggregate structures. It is a structure search system that searches.
- the material attribute information is used to select a new material, and the result of trying to synthesize with an actual synthesizer is used to update the database in real time, Even if the synthesis is successful, the data can be updated in real time with the results of measuring whether the material with the expected properties was produced, and the results are used to generate the feasibility model 25 and A physical property prediction model 24 is constructed. Therefore, it is possible to construct a material synthesis possibility prediction system and a measurement success probability prediction system by utilizing the feasibility database 23 that accumulates feasibility data (success or failure of synthesis and measurement) with matching material manufacturing conditions.
- the materials that can be selected in the database of the present disclosure are managed by descriptors, and the material property information of each material is structured to include at least two or more data items, and structured Some of the data items relate to material attributes.
- the attributes are, for example, attributes related to functional materials, such as low molecules, dyes, macromolecules, fluorescent/identifying isotope labels, self-organizing materials/structures, biomaterials (sugars, peptides, polypeptides, amino acids, proteins, fatty compounds, DNA, etc.), organic thin films (evaporation, coating processes), inorganic materials (solid phase method, coprecipitation method, melt quenching method, sol-gel method, etc.), nanoparticles, metal complexes, inorganic thin films (ALD , sputtering, etc.), synthetic materials based on synthetic biological techniques (genetic recombination and material synthesis using bacteria), crystal structures, nanostructures, microstructures, and the like.
- functional materials such as low molecules, dyes, macromolecules, fluorescent/identifying isotope labels, self-organizing materials/structures, biomaterials (sugars, peptides, polypeptides, amino acids, proteins, fatty compounds, DNA, etc.), organic thin films (
- the material searching and manufacturing system 1 has all properties and physical properties (mechanical properties, thermal properties, electrical properties, magnetic properties, optical properties, etc.) that can be evaluated by material production by physical, chemical, and biological methods , electrochemical properties, efficacy, toxicity, antibody response, interaction with cells, interaction with internal organs, intracellular transportability, in vivo transportability, adsorption, solubility, etc.). It's a common search system.
- Properties measured in the present disclosure include, for example, all properties and physical properties (mechanical properties, thermal properties, electrical properties, magnetic properties, optical properties, electrochemical properties, efficacy, toxicity, antibody response, interaction with cells, interaction with body organs, intracellular transportability, in vivo transportability, adsorptivity, solubility, etc.).
- a new material is selected based on information on all properties and physical properties that can be evaluated by material manufacturing by physical, chemical, and biological methods, and synthesis is performed by an actual synthesis apparatus.
- the database is updated in real time using the results of attempts, and even if the synthesis is successful, the data is updated in real time using the results of measuring whether materials with the expected properties were produced. and the results are used to build a feasibility model 25 and a property prediction model 24 . Therefore, it is possible to construct a new material manufacturing system utilizing the feasibility database 23 that stores feasibility data (success or failure of synthesis/measurement) with matching material manufacturing conditions.
- the material searching and manufacturing system 1 described above is merely an example of a processing system that performs various types of processing, and the processing system may be a system that is used for various purposes.
- the processing system may be a processing system used for the following applications.
- the feasibility database 23 containing both successful and unsuccessful data regarding material synthesis/physical property measurement can be obtained, it is possible to find out the characteristics of materials for successful material synthesis/physical property measurement and the characteristics of materials that fail. .
- the processing system is a search system that automatically presents the optimum equipment function expansion, such as automatic detection of problems and rate-limiting processes in the material manufacturing equipment, and indication of improvement policies, based on the analysis of success data and failure data.
- the processing system is a material search and manufacturing system that uses multiple material manufacturing equipment that can be accessed via the Internet depending on the situation. It is a universal feasibility prediction system by mutually exchanging and integrating feasibility data.
- an automatic material synthesis system that uses dedicated software to create a workflow for material synthesis, and in addition to device control such as material designation, dispensing amount, etc., and data management can be performed easily.
- analysis data during material production can be used to reflect in the database, and furthermore, the success or failure of synthesis in the most recent material synthesis, furthermore, the degree of goal achievement in the analysis results, Based on the above, it is possible to automatically select the next candidate material, so that it is possible to more efficiently search for a material that can select a material with a high probability of success.
- control network should be connected to the Internet via a wired network or a wireless network such as Wi-Fi (registered trademark) (Wireless-Fidelity), 4G/5G, or the like.
- Wi-Fi registered trademark
- 4G/5G Wireless-Fidelity
- synthesizer and the like used in the present disclosure are expected to be large-scale, but these can use virtualization technology on the cloud to integrally manage the distributed material manufacturing equipment as one system. By doing so, data can be transmitted and received in real time via the control interface.
- the next material search can be performed, and further, the material manufacturing apparatus can be continuously controlled to continue material synthesis.
- the feasibility database 23 of the present disclosure is a comparative and reusable feasibility database based on the standardization of feasibility data.
- the processing system is a highly accurate material development cost prediction system that utilizes the feasibility database 23 .
- the service provided by the processing system is a highly accurate materials search service utilizing feasibility model 25 .
- each component of each device illustrated is functionally conceptual and does not necessarily need to be physically configured as illustrated.
- the specific form of distribution/integration of each device is not limited to the illustrated one, and all or part of them can be functionally or physically distributed/integrated in arbitrary units according to various loads and usage conditions. Can be integrated and configured.
- FIG. 29 is a hardware configuration diagram showing an example of a computer that implements the functions of the information processing apparatus.
- the information processing apparatus 100 will be described below as an example.
- the computer 1000 has a CPU 1100 , a RAM 1200 , a ROM (Read Only Memory) 1300 , a HDD (Hard Disk Drive) 1400 , a communication interface 1500 and an input/output interface 1600 .
- Each part of computer 1000 is connected by bus 1050 .
- the CPU 1100 operates based on programs stored in the ROM 1300 or HDD 1400 and controls each section. For example, the CPU 1100 loads programs stored in the ROM 1300 or HDD 1400 into the RAM 1200 and executes processes corresponding to various programs.
- the ROM 1300 stores a boot program such as BIOS (Basic Input Output System) executed by the CPU 1100 when the computer 1000 is started, and programs dependent on the hardware of the computer 1000.
- BIOS Basic Input Output System
- the HDD 1400 is a computer-readable recording medium that non-temporarily records programs executed by the CPU 1100 and data used by such programs.
- the HDD 1400 is a recording medium that records an information processing program such as an information processing program according to the present disclosure, which is an example of the program data 1450 .
- a communication interface 1500 is an interface for connecting the computer 1000 to an external network 1550 (for example, the Internet).
- the CPU 1100 receives data from another device or transmits data generated by the CPU 1100 to another device via the communication interface 1500 .
- the input/output interface 1600 is an interface for connecting the input/output device 1650 and the computer 1000 .
- the CPU 1100 receives data from input devices such as a keyboard and mouse via the input/output interface 1600 .
- the CPU 1100 transmits data to an output device such as a display, speaker, or printer via the input/output interface 1600 .
- the input/output interface 1600 may function as a media interface for reading a program or the like recorded on a predetermined recording medium (media).
- Media include, for example, optical recording media such as DVD (Digital Versatile Disc) and PD (Phase change rewritable disk), magneto-optical recording media such as MO (Magneto-Optical disk), tape media, magnetic recording media, semiconductor memories, etc. is.
- the CPU 1100 of the computer 1000 implements the functions of the control unit 130 and the like by executing an information processing program such as an information processing program loaded on the RAM 1200 .
- the HDD 1400 also stores an information processing program such as an information processing program according to the present disclosure, and data in the storage unit 120 .
- CPU 1100 reads and executes program data 1450 from HDD 1400 , as another example, these programs may be obtained from another device via external network 1550 .
- the present technology can also take the following configuration.
- (3) The material selection is selected so as not to include materials that have been successfully synthesized in the past.
- the material selection is performed according to the priority based on the first predicted value of the material based on the material physical property information, The synthetic material selection method according to any one of (1) to (3).
- the first predicted value is output from a physical property prediction model in response to an input based on the material feature amount of each of two or more materials that constitute the material registered in the database, The synthetic material selection method according to (4).
- the material selection is performed according to a priority based on a second predicted value of materials based on the feasibility information;
- the second predicted value is selected from those above a predetermined threshold, The synthetic material selection method according to (6).
- the second predicted value is output from the feasibility model in response to input based on material feature values of each of two or more materials that constitute the material registered in the database, The synthetic material selection method according to (6) or (7).
- the selected materials are selected in the order of priority, and the selected materials are specified in a list format containing at least one item, wherein the item includes an identifier and a priority of the material.
- the synthetic material selection method according to any one of (4) to (8).
- the control device controls synthesis processing of the materials provided in the list format according to the priority, provides the synthesis processing result for each material, and the synthesis processing result is the controlled synthesis processing. material identifier and synthesis success/failure information,
- the control of the synthesis process includes a process of instructing a synthesizing device to perform the synthesizing process and receiving the synthesizing process result from the synthesizing device.
- the material physical property information is updated in the database with respect to the material that satisfies the target value, and based on the updated material physical property information, the physical property Including processing to update the forecast model, (13)
- the control device controls the indicated physical property measurement process for the successfully synthesized material identified by the identifier, and provides the identifier and the measured physical property value.
- the control of the physical property measurement process includes automatically moving the material synthesized by the synthesizing device to the measuring device, instructing the measuring device to perform the instructed physical property measurement, and performing the physical property measurement from the measuring device.
- the synthetic material selection method according to any one of (13) to (15).
- the material physical property information in the database is structured to include at least two items, one of the at least two items being an attribute item of the material.
- the synthetic material selection method according to any one of (1) to (16).
- a data structure used in an information processing system comprising a storage unit and a processing unit, stored in the storage unit, and used in the processing unit to select a material to be synthesized, ID for identifying materials, priority information indicating priority based on predicted physical property values of materials, feasibility information indicating feasibility of synthesizing materials, and synthesis success/failure information indicating success or failure of synthesis processing.
- the processing unit searches for a group of materials in which the synthesis success/failure information indicates unprocessed materials in descending order of priority based on the priority information, and the success or failure of the synthesis process indicated by the synthesis success/failure information is a predetermined number.
- a composite material selection data structure used in the process of selecting materials that meet the criteria.
- Material selection is performed according to the feasibility information output by the machine learning model, instructing a control device to synthesize the selected materials; updating information in the feasibility database based on the synthesis processing result from the control device; updating the machine learning model based on the updated database feasibility information; Manufacturing methods for manufacturing updated machine learning models.
- the attributes of the materials include low molecules, dyes, macromolecules, fluorescent/identifying isotope labels, self-organizing materials/structures, biomaterials (sugars, peptides, polypeptides, amino acids, proteins, fatty compounds, DNA, etc.), Organic thin films (evaporation, coating processes), inorganic materials (solid phase method, coprecipitation method, melt quenching method, sol-gel method, etc.), nanoparticles, metal complexes, inorganic thin films (ALD, sputtering, etc.), synthetic biological methods (including at least one of synthetic materials based on genetic recombination and material synthesis using bacteria, functional materials with crystal structures, nanostructures, and microstructures, The synthetic material selection method according to any one of (1) to (16).
- the processing indicated by the physical property measurement processing is a property that can be evaluated by material production by physical, chemical, or biological methods.
- the synthetic material selection method according to any one of (13) to (16).
- the properties include mechanical properties, thermal properties, electrical properties, magnetic properties, optical properties, electrochemical properties, efficacy, toxicity, antibody response, interaction with cells, interaction with internal organs, cell including at least one of internal transportability, in vivo transportability, adsorptivity, and solubility, (22)
- a processing unit that selects a material to be synthesized using material property information and feasibility information in a database, and instructs a control device to synthesize the selected material, and the control device a control unit that updates feasibility information in the database based on material synthesis processing results from
- a synthetic material selection device comprising: (25) Material selection for selecting materials to be synthesized using material property information and feasibility information in the database, instructing a control device to synthesize the selected materials; updating feasibility information in the database based on synthesis processing results from the control device; Synthetic material selection program that runs the process.
- Material search and manufacturing system (information processing system) REFERENCE SIGNS LIST 10 material manufacturing device 11 synthesis device 12 analysis device 13 measurement device 20 control device 21 control interface 22 physical property database 23 feasibility database 24 property prediction model 25 feasibility model 26 search agent N control network
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Chemical & Material Sciences (AREA)
- Computing Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Crystallography & Structural Chemistry (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Chemical Kinetics & Catalysis (AREA)
- Analytical Chemistry (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
Description
1.実施形態
1-1.本開示の実施形態に係る材料探索製造システムの構成及び処理の概要
1-1-1.材料探索製造システムの装置構成
1-1-2.材料探索製造システムの全体処理フロー
1-2.材料探索製造システムの各構成
1-2-1.材料製造装置
1-2-2.制御システム
1-2-2-1.データ
1-2-2-2.モデル
1-3.探索並びに材料製造処理後のデータベース及びモデルの更新処理例
1-3-1.物性予測モデルのみを用いた探索並びに材料製造処理後のデータベース及びモデルの更新
1-3-2.物性予測モデルと実行可能性モデルを用いた探索並びに材料製造処理後のデータベース及びモデルの更新
1-4.探索並びに材料製造処理後のデータベース及びモデルの更新処理の詳細
1-4-1.物性予測モデルのみを用いた探索並びに材料製造処理後のデータベース及びモデルの更新
1-4-2.物性予測モデルと実行可能性モデルを用いた探索並びに材料製造処理後のデータベース及びモデルの更新
1-5.実施例及び材料製造結果
1-6.実行性データベースの例
2.他の処理フロー例
2-1.合成に関する処理
2-2.測定に関する処理
2-3.システム構成例等
2-3-1.情報処理装置例
2-3-2.材料製造装置の構成例等
3.本開示の対象及び効果
4.その他
5.ハードウェア構成
[1-1.本開示の実施形態に係る材料探索製造システムの構成及び処理の概要]
以下では、まず、図1及び図2を用いて、以下に示す各種処理を行う情報処理システム(「処理システム」ともいう)の一例である材料探索製造システム1の構成、及び材料探索製造システム1が行う処理の概要について説明した後、各構成や処理の詳細について順を追って説明する。なお、以下では、材料探索製造システム1の構成及び処理に関する点を主に説明し、材料探索及び材料合成における従来技術の内容についての詳細な説明は適宜省略する。また、以下に示す材料探索製造システム1が処理の主体として記載されている処理については、材料探索製造システム1に含まれるいずれの装置が行ってもよい。
まず、図1に示す材料探索製造システム1の構成について説明する。図1は、本開示の材料探索製造システムの構成例を示す図である。図1に示すように、材料探索製造システム1には、材料製造装置10と、制御装置20と、外部記憶装置50とが含まれる。例えば、材料製造装置10と、制御装置20とは所定の通信網(図1では制御ネットワークN)を介して、有線または無線により通信可能に接続される。制御装置20と材料製造装置10とは、制御ネットワークNを通じて相互に情報のやり取りができる。制御ネットワークNは、無線、有線を介して通信できる専用ネットワーク、ローカルエリアネットワーク、インターネットあるいはクラウドと称される広域ネットワークのいずれであってもよい。
次に、図2を用いて材料探索製造システムの全体処理フローを説明する。図2は、材料探索製造システムによる処理手順を示すフローチャートである。例えば、図2は、材料探索製造システム1により行われる材料製造方法の一例を示す。例えば、材料探索製造システム1は、探索エージェント26により駆動され、次の手順通り材料探索を行う。
ここから、材料探索製造システム1の各構成の概要について説明する。
材料製造装置10に関する点について説明する。まず、材料製造装置10に関するハードウェアについて説明する。合成装置11は、混合・反応容器、試薬・溶媒ストック、生成物分離機能を内部に持つか、これら機能を持つ外部機器と接続されており、これら構成部は相互に内容物の受け渡しができる。また、合成装置11は、制御装置20からの命令に基づいて試薬・溶媒の混合、候補材料の合成、生成物の分離を行う。また、合成装置11はこれら操作が成功したかどうか検出部(検出用回路またはプロセッサ上で動作する検出用ソフトウェア・プログラム。以下同じ)を用いて判定を行い、判定結果を制御装置20に報告(送信)する。
次に、制御装置20に関する点について説明する。物性予測モデル24は、物性データベース22のデータから推定され、材料から物性値を予測する回帰モデルである。回帰モデルは、材料に関して予測値のみを返すものでも、予測値と予測値の分散を返すものでもよい。
ここから、制御装置20が用いるデータの例について図を参照しつつ説明する。
次に、制御装置20によるモデルを用いた処理の概要の一例について図を参照しつつ説明する。
次に、探索並びに材料製造処理後のデータベース及びモデルの更新処理の一例について説明する。
まず、物性予測モデル24のみを用いた処理の一例について図9を用いて説明する。図9は、物性予測モデルを用いた探索の一例を示す図である。図9に示す処理は、実行性データベース23に関わる処理以外は、図2に示したフローチャートの処理の流れに対応する。
次に、物性予測モデル24と実行可能性モデル25を用いた処理の一例について図10を用いて説明する。図10に示す処理は、図2に示したフローチャートの処理の流れに対応する。なお、図10において図9と同様の点については適宜説明を省略する。
ここで、上述した図9及び図10における処理の詳細を図示した図11~図15を用いて、各処理の詳細について説明する。
まず、図9に示した処理の詳細を図11及び図12を用いて説明する。図11及び図12は、物性予測モデル24を用いた処理の詳細例を示す図である。具体的には、図11は、図9における1回目~3回目の処理の詳細を示し、図12は、図9における4回目~6回目の詳細を示す。なお、図9と同様の点については詳細な説明を省略する。また、前出の通り、「成功」とは、合成の成功かつ測定の成功(図2のS7において「YES」)に該当することを意味し、「失敗」とは、合成の失敗(図2のS5において「NO」)または測定の失敗(図2のS7において「NO」)に該当することを意味する。
次に、図10に示した処理の詳細を図13~図15を用いて説明する。図13~図15は、物性予測モデル24及び実行可能性モデル25を用いた処理の詳細例を示す図である。例えば、図13~図15は、機械学習モデルである物性予測モデル24及び実行可能性モデル25を合成処理結果に基づいて製造する製造方法の一例を示す。具体的には、図13は、図10における1回目~2回目の処理の詳細を示し、図14は、図10における3回目~4回目の詳細を示し、図15は、図10における5回目~6回目の詳細を示す。なお、図10~図12と同様の点については詳細な説明を省略する。また、前出の通り、「成功」とは、合成の成功かつ測定の成功(図2のS7において「YES」)に該当することを意味し、「失敗」とは、合成の失敗(図2のS5において「NO」)または測定の失敗(図2のS7において「NO」)に該当することを意味する。
ここで、具体的な対象とするデータセットを用いた実施例を説明し、実施例を含む複数の手法の各々の材料製造結果について、図30及び図31を用いて説明する。図30及び図31は、探索に関する材料製造結果の一例を示す図である。
<手順#1>
手順#1-1:材料空間に含まれる未評価の分子からランダムに1分子を選択し、合成・測定を行う。当該分子を評価済みとする。
手順#1-2:合成・測定
(ア):合成・測定が失敗した場合:手順1から繰り返す。
(イ):合成・測定が成功した場合:分子構造と測定結果(物性値)を物性データベース22に保存する。
物性予測モデル24による探索(比較例#2)では、機械学習モデルとして予測平均値(μ)と予測分散(σ)を算出できるGaussian Process回帰を利用した、ベイズ最適化と呼ばれるアルゴリズムを実装した。機械学習モデルの入力には分子記述子(例えば、RdKit記述子)を用いた。探索の手順は下記の手順#2に示す通りである。
手順#2-1:すべての入力に対して等しい出力を与える物性予測モデル24(初期モデル)を構築する。
手順#2-2:物性予測モデル24を用いて、材料空間に含まれる未評価の分子に対して、次の定義で示される獲得関数(Acquisition Function:Fa)を計算する:
(ア):最大化問題のときは、式(9)を用いて獲得関数Faを計算
(イ):最小化問題のときは、式(10)を用いて獲得関数Faを計算
手順#2-4:最も優先順位の高い分子の合成・測定を行う。この分子を評価済みとする。
手順#2-5:合成・測定
(ア):合成・測定が失敗した場合:手順#2-4から繰り返す。
(イ):合成・測定が成功した場合:分子構造と測定結果(物性値)を物性データベース22に保存し、物性予測モデル24を更新する。
手順#2-6:物性データベース22にある測定値の最大値(最小値)が目標値に到達した場合は完了。まだの場合は手順#2-2から繰り返す。
物性予測モデル24のための機械学習モデルとしてGaussian Process回帰、実行可能性モデル25のための機械学習モデルとしてGaussian Process分類を利用した。機械学習モデルの入力は分子記述子(例えば、RdKit記述子)を用いた。探索の手順は下記の手順#3に示す通りである。例えば、実行可能性モデル25の初期モデルは、評価済みデータ(合成成功したデータ)を用いてモデル構築される。
手順#3-1:実行可能性モデル25(初期モデル)を、初期データ(データ数:100又は10)を用いて構築する。
手順#3-2:すべての入力に対して等しい出力を与える物性予測モデル24(初期モデル)を構築する。
手順#3-3:物性予測モデル24を用いて、材料空間に含まれる未評価の分子に対して、次の定義で示される獲得関数(Acquisition Function:Fa)を計算する:
(ア):最大化問題のときは、式(9)を用いて獲得関数Faを計算
(イ):最小化問題のときは、式(10)を用いて獲得関数Faを計算
手順#3-5:実行可能性モデル25によって、材料空間に含まれる未評価分子の実行可能性(合成・測定の成否の予測)を評価する。
手順#3-6:実行可能性が高く(確率0.5以上、等)、かつ最も優先順位の高い分子の合成・測定を行う。この分子を評価済みとする。
手順#3-7:合成・測定
(ア):合成・測定が失敗した場合:実行性データベース23および実行可能性モデル25を更新し、手順#3-5から繰り返す。
(イ):合成・測定が成功した場合:実行性データベース23および実行可能性モデル25の更新、分子構造と測定結果(物性値)から物性データベース22および物性予測モデル24を更新する。
手順#3-8:物性データベース22に保持された測定値の最大値(最小値)が目標値に到達した場合は完了。まだの場合は手順#3-2から繰り返す。
上述した実行性データベース23は一例に過ぎず、実行性データベース23には様々なデータ構造が採用されてもよい。この点について図16~図21を用いて説明する。図16~図21は、実行性データベース23の一例を示す図である。なお、上述した実行性データベース23と同様の点については適宜説明を省略する。
材料探索製造システム1は、上記した各種の情報を用いて任意の処理を実行可能である。この点について、上述した実施例とは別の実施例として以下説明する。材料探索製造システム1は、物性(特性)および/またはfeasibilityの情報に基づいて、priorityを計算し、priorityに従って、合成材料(分子)リスト(少なくとも一つを含む)を作成する。リストの各項目は、分子IDを少なくとも含む。なお、リストは、{分子ID、合成パスID}のペアを含んでも良い。
まず、合成に関する処理例について説明する。例えば、材料探索製造システム1は、以下の処理(1)~(8)の処理を実行する。なお、この合成処理は、探索エージェントが図2の処理を行う際に、「合成」処理(ステップS4)に対応して、図1の制御インターフェース21、合成装置11、及び分析装置12との間で行われる一連の処理の一例である。
(2)合成装置11は、リストの優先順位に従って、{分子ID(、合成パスID)}を読み出す。
(3)合成パスIDが指定されている場合には、合成装置11は、合成パスIDをキーとして、合成パスIDテーブルから反応ステップを取得し、合成処理を行う。
(4-1)合成装置11は、「合成不可」の通知を受けた場合、合成装置11は、制御インターフェース21に対し、最初の分子IDについて、合成失敗を通知し(図2の合成失敗)、(6)に進む。
(4-2)それ以外の場合、合成装置11は、反応ステップ1の処理を行う。処理の結果、「正常終了」または「異常終了」のいずれかの結果を通知する。
(4-3)合成装置11は、「異常終了」の場合、合成装置11は、制御インターフェース21に対し、最初の分子IDについて、合成失敗を通知し(図2の合成失敗)、(6)に進む。
(4-4)合成装置11は、「正常終了」の場合、次の反応ステップがあるかどうか確認し、次の反応ステップがある場合には、次の反応ステップを取得し、(4)から繰り返す。次の反応ステップが無い場合には、(5)に進む。
(5-1)分析装置12が、合成成功と判断する場合、合成装置11に「合成成功」を通知。それ以外の場合、「合成失敗」を通知する。
(6-1)合成不可、異常終了、合成失敗の場合、「合成失敗」を送信する。
(6-2)合成成功の場合、「合成成功」を送信する。
(8)上記合成装置11の処理と並列に、探索エージェント26は、「合成失敗」の場合、合成ができないと判断し、「実行可能性モデル」の構築、または更新を行う。
次に、測定に関する処理例について説明する。例えば、材料探索製造システム1は、以下の処理(11)~(16)の処理を実行する。なお、この測定処理は、探索エージェントが図2の処理を行う際に、「測定」処理(ステップS6)に対応して、図1の制御インターフェース21、合成装置11、及び測定装置13との間で行われる一連の処理の一例である。
(12)合成の成功した分子IDを特定し、測定すべき物性を指定して、制御インターフェース21を経由して、合成装置11に対して、合成の成功した材料の測定指示を行う。
(13)指示を受けた合成装置11は、指示された分子IDで特定される合成材料を測定装置13に提供するように自動制御し、測定装置13に対して、提供した材料について指定された物性の測定開始を指示する。
(15)探索エージェント26は、測定結果が、「測定不可」の場合、合成ができないと判断し、「実行可能性モデル」の構築、または更新を行う。
なお、上記の例では、合成フェーズと測定フェーズを分け、それぞれの指示を、探索エージェント26が、合成エージェントに行うようにしたが、探索エージェント26の合成の指示の後、合成装置11が、合成処理の「分析結果を通知」を行った後、自動的に、測定処理の「測定を指示」を行うようにしてもよい。また、上記の例では、制御インターフェース21が、情報の変換を行っているが、合成装置11が、情報変換を行ってもよい。
ここで、情報処理装置100の構成例について説明する。図25は、本開示の情報処理装置の構成例を示す図である。例えば、情報処理装置100は、制御装置20として機能する。情報処理装置100は、制御装置200と一体であってもよい。
ここで、材料製造装置10において、合成装置11と分析装置12または測定装置13との間で自動的に合成材料を移動するための機構(「材料移動機構」ともいう)を有することが望ましい。この点について、以下図26~図28を用いて説明する。図26~図28は、材料製造装置の構成の一例を示す図である。なお、以下では、材料移動機構の態様に応じて、材料製造装置10A、10B、10Cとして説明するが、材料移動機構に関する構成以外については、材料製造装置10と同様である。また、材料製造装置10について、上述した内容と同様の点については適宜説明を省略する。
上記のように、本開示に係る合成材料選択方法は、データベースの材料物性情報(上記例ではデータベースDB1に示す情報等。以下同様)と実行可能性情報(上記例ではデータベースDB2に示す情報等。以下同様)を用いて合成対象の材料を選択する材料選択を行い、制御装置(上記例では材料製造装置10、制御装置20、情報処理装置100または制御装置200等。以下同様)に対して、選択された材料の合成処理の指示を行い、制御装置からの合成処理結果に基づいて、データベースの実行可能性情報を更新する。このように、例えば、材料探索製造システム1が実行する合成材料選択方法は、データベースの材料物性情報と実行可能性情報を用いて合成対象の材料を選択する材料選択を行うことと、選択された材料の合成処理の指示を行い、合成処理結果に基づいて、データベースの実行可能性情報を更新すること、とを含む。これにより、合成材料選択方法は、合成の実行性に基づく材料合成を可能にすることができる。
上記各実施形態において説明した各処理のうち、自動的に行われるものとして説明した処理の全部または一部を手動的に行うこともでき、あるいは、手動的に行われるものとして説明した処理の全部または一部を公知の方法で自動的に行うこともできる。この他、上記文書中や図面中で示した処理手順、具体的名称、各種のデータやパラメータを含む情報については、特記する場合を除いて任意に変更することができる。例えば、各図に示した各種情報は、図示した情報に限られない。
上述してきた各実施形態や変形例に係る情報処理装置100、制御装置200、材料製造装置10等の情報機器は、例えば図29に示すような構成のコンピュータ1000によって実現される。図29は、情報処理装置の機能を実現するコンピュータの一例を示すハードウェア構成図である。以下、情報処理装置100を例に挙げて説明する。コンピュータ1000は、CPU1100、RAM1200、ROM(Read Only Memory)1300、HDD(Hard Disk Drive)1400、通信インターフェース1500、及び入出力インターフェース1600を有する。コンピュータ1000の各部は、バス1050によって接続される。
(1)
データベースの材料物性情報と実行可能性情報に基づいて合成対象の材料を選択する材料選択を行い、
制御装置に対して、前記選択された材料の合成処理の指示を行い、
前記制御装置からの合成処理結果に基づいて、前記データベースの実行可能性情報を更新する、
合成材料選択方法。
(2)
前記データベースは、二以上の合成処理対象の材料を含む、
(1)に記載の合成材料選択方法。
(3)
前記材料選択では、過去に合成の成功した材料を含まないように選択される、
(1)または(2)に記載の合成材料選択方法。
(4)
前記材料選択では、前記材料物性情報に基づく材料の第一の予測値に基づく優先順位に従って行われる、
(1)~(3)のいずれか1つに記載の合成材料選択方法。
(5)
前記第一の予測値は、前記データベースに登録された材料を構成する二以上の材料各々の材料特徴量に基づく入力に応じて、物性予測モデルから出力される、
(4)に記載の合成材料選択方法。
(6)
前記材料選択では、前記実行可能性情報に基づく材料の第二の予測値に基づく優先順位に従って行われる、
(4)または(5)に記載の合成材料選択方法。
(7)
前記材料選択では、第二の予測値が所定の閾値以上のものから選択される、
(6)に記載の合成材料選択方法。
(8)
前記第二の予測値は、前記データベースに登録された材料を構成する二以上の材料各々の材料特徴量に基づく入力に応じて実行可能性モデルから出力される、
(6)または(7)に記載の合成材料選択方法。
(9)
前記選択された材料は、前記優先順位の順に選択され、選択された材料は、少なくとも一つの項目を含むリスト形式で指定され、前記項目は、前記材料の識別子及び優先順位を含む、
(4)~(8)のいずれか1つに記載の合成材料選択方法。
(10)
前記制御装置は、前記優先順位に従って前記リスト形式で提供される前記材料の合成処理を制御し、前記材料ごとに前記合成処理結果を提供し、前記合成処理結果は、前記制御された合成処理の材料の識別子及び合成の成否情報である、
(9)に記載の合成材料選択方法。
(11)
前記合成処理の制御は、合成装置に合成処理を指示し、前記合成装置から、前記合成処理結果を受信する処理を含む、
(10)に記載の合成材料選択方法。
(12)
前記実行可能性情報の更新では、前記合成の成否情報が成功の場合には、前記データベースにおいて、前記合成の成功した材料について前記実行可能性情報を更新し、さらに更新した実行可能性情報に基づき、実行可能性モデルを更新する、
(11)に記載の合成材料選択方法。
(13)
前記合成の成否情報に基づいて、前記制御装置に対し、前記合成の成功した材料の識別子及び測定する物性を指定して、物性測定処理の指示を行い、前記制御装置からの物性測定処理結果に基づいて、前記データベースにおける前記合成の成功した材料の材料物性情報を更新する処理を含む、
(11)または(12)に記載の合成材料選択方法。
(14)
前記材料物性情報の更新は、前記物性測定処理結果が目標値を満たす場合には、前記データベースにおいて、前記目標値を満たす材料について前記材料物性情報を更新し、更新した材料物性情報に基づき、物性予測モデルを更新する処理を含む、
(13)に記載の合成材料選択方法。
(15)
前記制御装置は、前記識別子で特定される前記合成の成功した材料に対する、前記指示された物性測定処理の制御を行い、前記識別子及び前記測定された物性値を提供する、
(13)または(14)に記載の合成材料選択方法。
(16)
前記物性測定処理の制御は、前記合成装置の合成した材料を自動的に測定装置に移動し、前記指示された物性測定を行うことを前記測定装置に指示し、前記測定装置から、前記物性測定処理結果を受信する処理を含む、
(13)~(15)のいずれか1つに記載の合成材料選択方法。
(17)
前記データベースの材料物性情報は少なくとも二の項目を含むように構造化されており、前記少なくとも二の項目のうちの一は、前記材料の属性項目である、
(1)~(16)のいずれか1つに記載の合成材料選択方法。
(18)
データベースの材料物性情報と実行可能性情報を用いて合成対象の材料を選択する材料選択を行い、
制御装置に対して、前記選択された材料の合成処理の指示を行い、
前記制御装置からの合成処理結果に基づいて、前記データベースの実行可能性情報を更新する、
処理により材料を製造する材料製造方法。
(19)
記憶部及び処理部を備える情報処理システムに用いられ、前記記憶部に記憶され、前記処理部における合成対象材料の選択に用いられるデータ構造であって、
材料を識別するIDと、予測される材料の物性値に基づく優先順位を示す優先順位情報と、材料の合成の実行可能性を示す実行可能性情報と、合成処理の成否を示す合成成否情報とを含み、
前記処理部が、前記合成成否情報が未処理を示す材料群を対象として、前記優先順位情報に基づく優先順位が高い方から順に探索し、前記合成成否情報が示す前記合成処理の成否が所定の基準を満たす材料を選択する処理に用いられる合成材料選択データ構造。
(20)
機械学習モデルの出力する実行可能性情報に応じて材料選択を行い、
制御装置に対して、前記選択された材料の合成処理の指示を行い、
前記制御装置からの合成処理結果に基づいて、実行性データベースの情報を更新し、
前記更新されたデータベースの実行可能性情報に基づいて前記機械学習モデルを更新し、
更新された機械学習モデルを製造する製造方法。
(21)
前記材料の属性は、低分子、色素、高分子、蛍光・同定同位体標識、自己組織化材料・構造体、生体材料(糖類、ペプチド、ポリペプチド、アミノ酸、タンパク質、脂肪化合物、DNA等)、有機薄膜(蒸着、塗布プロセス)、無機材料(固相法、共沈法、溶融急冷法、ゾルゲル法等)、ナノ粒子、金属錯体、無機薄膜(ALD、スパッタリング等)、合成生物学的手法(遺伝子組み換えと細菌利用による材料合成)に基づく合成材料、結晶構造、ナノ構造、マイクロ構造を持つ機能性材料のうち、少なくとも一つを含む、
(1)~(16)のいずれか1つに記載の合成材料選択方法。
(22)
前記物性測定処理で指示される処理は、物理・化学・生物的な方法による材料製造によって評価できる特性である、
(13)~(16)のいずれか1つに記載の合成材料選択方法。
(23)
前記特性は、機械的性質、熱的性質、電気的性質、磁気的性質、光学的性質、電気化学的特性、薬効、毒性、抗体反応、細胞との相互作用、体内器官との相互作用、細胞内の輸送性、生体内輸送性、吸着性、溶解性のうち、少なくとも一つを含む、
(22)に記載の合成材料選択方法。
(24)
データベースの材料物性情報と実行可能性情報を用いて合成対象の材料を選択する材料選択を行う処理部と、制御装置に対して、前記選択された材料の合成処理の指示を行い、前記制御装置からの材料合成処理結果に基づいて、前記データベースの実行可能性情報を更新する制御部、
を備える合成材料選択装置。
(25)
データベースの材料物性情報と実行可能性情報を用いて合成対象の材料を選択する材料選択を行い、
制御装置に対して、前記選択された材料の合成処理の指示を行い、
前記制御装置からの合成処理結果に基づいて、前記データベースの実行可能性情報を更新する、
処理を実行させる合成材料選択プログラム。
10 材料製造装置
11 合成装置
12 分析装置
13 測定装置
20 制御装置
21 制御インターフェース
22 物性データベース
23 実行性データベース
24 物性予測モデル
25 実行可能性モデル
26 探索エージェント
N 制御ネットワーク
Claims (20)
- データベースの材料物性情報と実行可能性情報に基づいて合成対象の材料を選択する材料選択を行い、
制御装置に対して、前記選択された材料の合成処理の指示を行い、
前記制御装置からの合成処理結果に基づいて、前記データベースの実行可能性情報を更新する、
合成材料選択方法。 - 前記データベースは、二以上の合成処理対象の材料を含む、
請求項1に記載の合成材料選択方法。 - 前記材料選択では、過去に合成の成功した材料を含まないように選択される、
請求項1に記載の合成材料選択方法。 - 前記材料選択では、前記材料物性情報に基づく材料の第一の予測値に基づく優先順位に従って行われる、
請求項1に記載の合成材料選択方法。 - 前記第一の予測値は、前記データベースに登録された材料を構成する二以上の材料各々の材料特徴量に基づく入力に応じて物性予測モデルから出力される、
請求項4に記載の合成材料選択方法。 - 前記材料選択では、前記実行可能性情報に基づく材料の第二の予測値に基づく優先順位に従って行われる、
請求項4に記載の合成材料選択方法。 - 前記材料選択では、第二の予測値が所定の閾値以上のものから選択される、
請求項6に記載の合成材料選択方法。 - 前記第二の予測値は、前記データベースに登録された材料を構成する二以上の材料各々の材料特徴量に基づく入力に応じて実行可能性モデルから出力される、
請求項6に記載の合成材料選択方法。 - 前記選択された材料は、前記優先順位の順に選択され、選択された材料は、少なくとも一つの項目を含むリスト形式で指定され、前記項目は、前記材料の識別子及び優先順位を含む、
請求項4に記載の合成材料選択方法。 - 前記制御装置は、前記優先順位に従って前記リスト形式で提供される前記材料の合成処理を制御し、前記材料ごとに前記合成処理結果を提供し、前記合成処理結果は、前記制御された合成処理の材料の識別子及び合成の成否情報である、
請求項9に記載の合成材料選択方法。 - 前記合成処理の制御は、合成装置に合成処理を指示し、前記合成装置から、前記合成処理結果を受信する処理を含む、
請求項10に記載の合成材料選択方法。 - 前記実行可能性情報の更新では、前記合成の成否情報が成功の場合には、前記データベースにおいて、前記合成の成功した材料について前記実行可能性情報を更新し、さらに更新した実行可能性情報に基づき、実行可能性モデルを更新する、
請求項11に記載の合成材料選択方法。 - 前記合成の成否情報に基づいて、前記制御装置に対し、前記合成の成功した材料の識別子及び測定する物性を指定して、物性測定処理の指示を行い、前記制御装置からの物性測定処理結果に基づいて、前記データベースにおける前記合成の成功した材料の材料物性情報を更新する処理を含む、
請求項11に記載の合成材料選択方法。 - 前記材料物性情報の更新は、前記物性測定処理結果が目標値を満たす場合には、前記データベースにおいて、前記目標値を満たす材料について前記材料物性情報を更新し、更新した材料物性情報に基づき、物性予測モデルを更新する処理を含む、
請求項13に記載の合成材料選択方法。 - 前記制御装置は、前記識別子で特定される前記合成の成功した材料に対する、前記指示された物性測定処理の制御を行い、前記識別子及び前記測定された物性値を提供する、
請求項13に記載の合成材料選択方法。 - 前記物性測定処理の制御は、前記合成装置の合成した材料を自動的に測定装置に移動し、前記指示された物性測定を行うことを前記測定装置に指示し、前記測定装置から、前記物性測定処理結果を受信する処理を含む、
請求項13に記載の合成材料選択方法。 - 前記データベースの材料物性情報は少なくとも二の項目を含むように構造化されており、前記少なくとも二の項目のうちの一は、前記材料の属性項目である、
請求項1に記載の合成材料選択方法。 - データベースの材料物性情報と実行可能性情報を用いて合成対象の材料を選択する材料選択を行い、
制御装置に対して、前記選択された材料の合成処理の指示を行い、
前記制御装置からの合成処理結果に基づいて、前記データベースの実行可能性情報を更新する、
処理により材料を製造する材料製造方法。 - 記憶部及び処理部を備える情報処理システムに用いられ、前記記憶部に記憶され、前記処理部における合成対象材料の選択に用いられるデータ構造であって、
材料を識別するIDと、予測される材料の物性値に基づく優先順位を示す優先順位情報と、材料の合成の実行可能性を示す実行可能性情報と、合成処理の成否を示す合成成否情報とを含み、
前記処理部が、前記合成成否情報が未処理を示す材料群を対象として、前記優先順位情報に基づく優先順位が高い方から順に探索し、前記合成成否情報が示す前記合成処理の成否が所定の基準を満たす材料を選択する処理に用いられる合成材料選択データ構造。 - 機械学習モデルの出力する実行可能性情報に応じて材料選択を行い、
制御装置に対して、前記選択された材料の合成処理の指示を行い、
前記制御装置からの合成処理結果に基づいて、実行性データベースの情報を更新し、
前記更新されたデータベースの実行可能性情報に基づいて前記機械学習モデルを更新し、
更新された機械学習モデルを製造する製造方法。
Priority Applications (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
EP22886869.1A EP4428866A1 (en) | 2021-11-01 | 2022-10-20 | Synthesis material selection method, material production method, synthesis material selection data structure, and production method |
CN202280071623.0A CN118160042A (zh) | 2021-11-01 | 2022-10-20 | 合成材料选择方法、材料制造方法、合成材料选择数据结构和制造方法 |
JP2023556385A JPWO2023074542A1 (ja) | 2021-11-01 | 2022-10-20 |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2021-178996 | 2021-11-01 | ||
JP2021178996 | 2021-11-01 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2023074542A1 true WO2023074542A1 (ja) | 2023-05-04 |
Family
ID=86159416
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2022/039200 WO2023074542A1 (ja) | 2021-11-01 | 2022-10-20 | 合成材料選択方法、材料製造方法、合成材料選択データ構造、及び製造方法 |
Country Status (4)
Country | Link |
---|---|
EP (1) | EP4428866A1 (ja) |
JP (1) | JPWO2023074542A1 (ja) |
CN (1) | CN118160042A (ja) |
WO (1) | WO2023074542A1 (ja) |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2003014728A (ja) * | 2000-06-14 | 2003-01-15 | Pfizer Pharmaceuticals Inc | 薬物動態学的物性の予測方法及びシステム |
JP2005166053A (ja) * | 2003-12-01 | 2005-06-23 | Accelrys Software Inc | 高速処理の実験情報をデータベースに保存する方法 |
WO2017221444A1 (ja) | 2016-06-21 | 2017-12-28 | 国立研究開発法人物質・材料研究機構 | 探索システム、探索方法および物性データベース管理装置 |
JP2021140701A (ja) * | 2020-03-09 | 2021-09-16 | 株式会社豊田中央研究所 | 材料設計プログラム |
-
2022
- 2022-10-20 CN CN202280071623.0A patent/CN118160042A/zh active Pending
- 2022-10-20 EP EP22886869.1A patent/EP4428866A1/en active Pending
- 2022-10-20 WO PCT/JP2022/039200 patent/WO2023074542A1/ja active Application Filing
- 2022-10-20 JP JP2023556385A patent/JPWO2023074542A1/ja active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2003014728A (ja) * | 2000-06-14 | 2003-01-15 | Pfizer Pharmaceuticals Inc | 薬物動態学的物性の予測方法及びシステム |
JP2005166053A (ja) * | 2003-12-01 | 2005-06-23 | Accelrys Software Inc | 高速処理の実験情報をデータベースに保存する方法 |
WO2017221444A1 (ja) | 2016-06-21 | 2017-12-28 | 国立研究開発法人物質・材料研究機構 | 探索システム、探索方法および物性データベース管理装置 |
JP2021140701A (ja) * | 2020-03-09 | 2021-09-16 | 株式会社豊田中央研究所 | 材料設計プログラム |
Non-Patent Citations (3)
Title |
---|
HATANAKA, MIHO: "Achievements and challenges of materials informatics in chemistry", JOURNAL OF JAPANESE SOCIETY FOR ARTIFICIAL INTELLIGENCE, vol. 34, no. 3, 1 May 2019 (2019-05-01), pages 351 - 357, XP009545490, ISSN: 2188-2266, DOI: 10.11517/jjsai.34.3_351 * |
HISAKI IKEBATAKENTA HONGOTETSU ISOMURARYO MAEZONORYO YOSHIDA: "Bayesian molecular design with a chemical language model", JOURNAL OF COMPUTER-AIDED MOLECULAR DESIGN, vol. 31, 2017, pages 379 - 391, XP036211747, DOI: 10.1007/s10822-016-0008-z |
KIM, Y.KIM, E.ANTONO, E. ET AL.: "Machine-learned metrics for predicting the likelihood of success in materials discovery", NPJ COMPUT MATER, vol. 6, 2020, pages 131 |
Also Published As
Publication number | Publication date |
---|---|
EP4428866A1 (en) | 2024-09-11 |
JPWO2023074542A1 (ja) | 2023-05-04 |
CN118160042A (zh) | 2024-06-07 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Strieth-Kalthoff et al. | Machine learning the ropes: principles, applications and directions in synthetic chemistry | |
WO2019186194A2 (en) | Ensemble model creation and selection | |
Tsirigos et al. | PRED-TMBB2: improved topology prediction and detection of beta-barrel outer membrane proteins | |
Yeang et al. | Physical network models | |
Amabilino et al. | Guidelines for recurrent neural network transfer learning-based molecular generation of focused libraries | |
US20210090690A1 (en) | Molecular design using reinforcement learning | |
KR20220092881A (ko) | 목적하는 유기 분자에 대한 유기 합성 경로를 설계하기 위한 시스템 및 방법 | |
Harada | Simple, yet efficient conformational sampling methods for reproducing/predicting biologically rare events of proteins | |
Powers et al. | Fragment-based ligand generation guided by geometric deep learning on protein-ligand structure | |
US11215467B1 (en) | Method of and system for path selection | |
JP5211458B2 (ja) | 化合物の仮想スクリーニング方法および装置 | |
Moret et al. | Perplexity-based molecule ranking and bias estimation of chemical language models | |
Hu et al. | De novo drug design based on Stack-RNN with multi-objective reward-weighted sum and reinforcement learning | |
Stringer et al. | PIPENN: protein interface prediction from sequence with an ensemble of neural nets | |
US12018948B2 (en) | Method of and system for path selection | |
Wang et al. | A transformer-based generative model for de novo molecular design | |
Wu et al. | Spatial graph attention and curiosity-driven policy for antiviral drug discovery | |
WO2023074542A1 (ja) | 合成材料選択方法、材料製造方法、合成材料選択データ構造、及び製造方法 | |
Kesheri et al. | Advances in soft computing approaches for gene prediction: a bioinformatics approach | |
Khemchandani et al. | DeepGraphMol, a multi-objective, computational strategy for generating molecules with desirable properties: a graph convolution and reinforcement learning approach | |
Wang et al. | A pre-trained conditional transformer for target-specific de novo molecular generation | |
Chin et al. | Predicting condensate formation of protein and RNA under various environmental conditions | |
Ghandikota et al. | Multimodal feature learning framework for disease biomarker discovery | |
Gambacorta et al. | Machine learning resources for drug design | |
JP7495549B1 (ja) | 物質の探索支援方法、物質の探索支援装置、コンピュータプログラム及び物質の製造方法 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 22886869 Country of ref document: EP Kind code of ref document: A1 |
|
WWE | Wipo information: entry into national phase |
Ref document number: 2023556385 Country of ref document: JP |
|
WWE | Wipo information: entry into national phase |
Ref document number: 18703731 Country of ref document: US |
|
WWE | Wipo information: entry into national phase |
Ref document number: 202280071623.0 Country of ref document: CN |
|
WWE | Wipo information: entry into national phase |
Ref document number: 2022886869 Country of ref document: EP |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
ENP | Entry into the national phase |
Ref document number: 2022886869 Country of ref document: EP Effective date: 20240603 |