CN102930113A - Building method of two-level fitting quantitative structure-activity relationship (QSAR) model for forecasting compound activity - Google Patents
Building method of two-level fitting quantitative structure-activity relationship (QSAR) model for forecasting compound activity Download PDFInfo
- Publication number
- CN102930113A CN102930113A CN2012104552399A CN201210455239A CN102930113A CN 102930113 A CN102930113 A CN 102930113A CN 2012104552399 A CN2012104552399 A CN 2012104552399A CN 201210455239 A CN201210455239 A CN 201210455239A CN 102930113 A CN102930113 A CN 102930113A
- Authority
- CN
- China
- Prior art keywords
- compound
- model
- activity
- training set
- qsar
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Landscapes
- Investigating Or Analysing Biological Materials (AREA)
Abstract
The invention discloses a building method of a two-level fitting quantitative structure-activity relationship (QSAR) model for forecasting compound activity. The building method includes following procedures: 1 a plurality of compounds with the same frames are utilized as a training set, and the train set compounds are divided into substituent groups and are coincided; 2 a linear regression method is utilized to calculate local physiological action produced by each substituent group, and a preceding-stage fitting model is built; 3 according to the local physiological action which is obtained in calculating mode in the procedure 2, a neural network method is utilized to calculate the whole biological activity, and a backward-stage fitting model is built; and 4 the preceding-stage fitting model and the backward-stage fitting model are combined to form the front-and-back two-stage QSAR model. According to the building method, the linear regression method and the neural network method are combined to build the model, the neural network method has good fitting performance, and compared with a traditional linear model, the built model can accurately forecast the biological activity of the compounds.
Description
Technical field
The present invention relates to a kind of construction method of OSAR model, especially a kind of construction method of the two-stage match QSAR model for the predictive compound activity belongs to the biological medicine areas of information technology.
Background technology
Quantitative structure activity relationship (Quantitative Structure-Activity Relationship is called for short QSAR) is a kind of technology by mathematical model quantitative forecast compound activity.Because the result of study of 3D QSAR has clear and definite directive significance, is extensively adopted by many researchs at present.But because the modeling process of 3D QSAR is carried out in the black box of business software, and the process in the software black box is difficult to human intervention, this has increased the difficulty of its modeling optimization undoubtedly, and a kind of modeling method of publishing, generally acknowledging convenient and swift 3D QSAR is not yet arranged so far.Therefore, it is significant to set up a kind of conveniently 3D QSAR modeling method.
At present, the 3D QSAR method of putting down in writing at publication is in modeling process, not only compound is superimposed irregular, and use traditional linear regression method (such as partial least square method etc.), in the process of model of fit, only consider the complicacy that organic chemistry is theoretical, do not consider biological acceptor, cause not meeting biological chemistry theory, the final goodness of fit and the predictive ability of impact.
Topomer composite technology based on bee-line has neat superimposed result, is a kind of compound folding method with optimistic application prospect.If can consider the complicacy of biological acceptor, so that meet the biological chemistry theory based on the QSAR result of study of Topomer folding method, then can improve the goodness of fit and the predictive ability of QSAR model.
Neural network (Neural Networks) is the statistical modeling method that a kind of physiological function by simulation mammal brain is carried out data fitting.Neural network model successfully has been applied to the function prediction of biomacromolecule, the toxicity prediction of organic contaminant, the performance prediction of macromolecule polymeric material etc., and the application in the chemicals MOLECULE DESIGN is also with increasingly extensive.Because neural network is approached the mapping relations of any complexity, therefore divide greatly the period of the day from 11 p.m. to 1 a.m for the biological acceptor more complicated than little molecule when the function of chemical compound target, can be than the linear model biologically active of predictive compound more accurately based on the QSAR model of neural network.
QSAR modeling based on neural network generally needs by following three steps: 1) activity data of arrangement compound is as dependent variable; 2) select suitable descriptor as independent variable and calculating; 3) select suitable neural net method to make up the QSAR model.
Wherein, selecting suitable descriptor is the necessary condition of setting up the neural network QSAR model with good predict ability as independent variable.If the information gain that independent variable contains is not enough, then institute's established model is difficult to have good predictive ability, although yet the number that increases independent variable might improve information gain, over-fitting, Convergent Phenomenon and cause model performance to descend even the modeling failure not can appear.Therefore, seek a kind of low dimensional vector that comprises the enough information gain as independent variable, very crucial based on the QSAR model of neural network for making up.
Summary of the invention
Purpose of the present invention is in order to solve the defective of above-mentioned prior art, and a kind of construction method with the bioactive two-stage match QSAR model for the predictive compound activity of good fit goodness, Accurate Prediction compound is provided.
Purpose of the present invention can reach by taking following technical scheme:
Be used for the construction method of the two-stage match QSAR model of predictive compound activity, it is characterized in that may further comprise the steps:
1) get several compounds with identical skeleton as training set, the training set compound is divided substituting group, and superimposed training set compound;
2) according to the structure and activity of training set compound, adopt linear regression method to calculate the local physiological action that each substituting group produces, set up the prime model of fit;
3) according to activity and the step 2 of training set compound) the local physiological action that calculates, adopt neural network to calculate the whole biologically active of compound, set up the rear class model of fit;
4) with prime model of fit and the combination of rear class model of fit, be built into front and back stages QSAR model.
As a kind of preferred version, step 2) activity of described training set compound is inhibition concentration or inhibiting rate.
As a kind of preferred version, step 1) is specific as follows:
For existing compound, carry out the data acquisition of biologic activity for the particular test system, data target adopts the negative logarithmic form ［ lg (inhibition concentration) or-lg (1/ inhibiting rate-1) ］ of inhibition concentration or inhibiting rate, with this as the training set sample; Use the two-dimensional structure of Sybyl analysis software check compound, the compound by check is generated its three-dimensional structure; Subsequently, the substituting group of Further Division compound, and be optimized; At last, divide based on substituting group, and adopt the Topomer composite technology to carry out superimposed to above compound three-dimensional structure.
As a kind of preferred version, step 2) specific as follows:
With the molecular field around the superimposed training set compound of probe scanning, calculate MSA, CoMFA or CoMSIA molecular field, after molecular field information selected, carry out linear regression with the experiment activity of training set compound, obtain the prime model of fit of structure-activity relationship.
As a kind of preferred version, step 3) is specific as follows:
With step 2) the local physiological action that calculates, carry out normalization with the activity of training set compound, obtain normalized value, go normalization by neural network model, calculate the whole biologically active of compound, obtain the rear class model of fit.
As a kind of preferred version, described training set compound is the pyrazole compound with p38 kinase inhibition rate.
As a kind of preferred version, the sample size of described training set compound has 30 at least.
As a kind of preferred version, the substituting group that described training set compound is divided has 2 at least, includes the connecting bridge of compound in the substituting group of described division.
As a kind of preferred version, described step 2) linear regression method that adopts is partial least square method or principal component analysis (PCA).
As a kind of preferred version, the neural network that described step 3) adopts is BF neural network or RBF neural network
The present invention has following beneficial effect with respect to prior art:
1, modeling method of the present invention is to have adopted the mode of linear regression method and neural network combination to set up model, and because neural network has good capability of fitting, the model of structure can be than the conventional linear model biologically active of predictive compound more accurately.
2, modeling method of the present invention adopts linear regression to have two aspect beneficial effects as the prime model: 1) linear model is explained easily, helps the structural modification of compound; 2) with the result of the prime model independent variable as the rear class neural network model, can avoid occurring not restraining, the phenomenon of over-fitting; Thereby prevent neural net model establishing failure, improve the predictive ability of rear class model, namely improve the predictive ability of whole front and back stages model of fit.
3, to have adopted the Topomer composite technology that the training set compound is carried out superimposed for modeling method of the present invention, is conducive to the efficient of modeling, and simultaneously superimposed result is neat.
4, modeling method of the present invention need not molecular docking, need not quantum chemistry calculation, the independent variable number of neural network is few, can within the identical time, can make up based on the training set of large sample like this and obtain model, thereby can further improve the predictive ability of QSAR model.
5, modeling method of the present invention has solved and has used conventional linear to return the coarse problem that predicts the outcome that causes as not the considering the biological acceptor complicacy of modeling method, the two-stage match QSAR model that makes up is to the p38 kinase inhibiting activity of pyrazole compound, related coefficient is square greater than 0.95, present good good capability of fitting and estimated performance, have broad application prospects as the biologically active Forecasting Methodology of the pyrazoles immunosuppressive drug take the p38 kinases as action target spot, anti-inflammatory agent, antifungal.
Description of drawings
Fig. 1 is the schematic flow sheet of two-stage match QSAR model building method of the present invention.
Fig. 2 is that pyrazoles p38 inhibitors of kinases training set compound adopts the traditional single stage model M
1Goodness of fit scatter diagram.
Fig. 3 is that pyrazoles p38 inhibitors of kinases training set compound adopts the front and back stages model M
1-M
2Goodness of fit scatter diagram.
Fig. 4 is that pyrazoles p38 inhibitors of kinases training set compound adopts the traditional single stage model M
1The estimated performance scatter diagram.
Fig. 5 is that pyrazoles p38 inhibitors of kinases training set compound adopts the front and back stages model M
1-M
2The estimated performance scatter diagram.
Embodiment
Embodiment 1:
As shown in Figure 1, the linear regression of present embodiment-neural network front and back stages match QSAR model, its construction step is as follows:
1) bioactive arrangement
For guaranteeing the statistics effect, get 35 pyrazole compounds with p38 kinase inhibition rate as training set S
1, its inhibiting rate α is converted into logarithmic form: Y
1=LgBio=-lg (α
-1-1).Y
1=LgBio is the used dependent variable of follow-up modeling, uses the two-dimensional structure of Sybyl analysis software check compound, and the compound by check is generated its three-dimensional structure.
2) structure of prime model of fit
With the training set compound S
1Import the molecule list S1.tbl of Sybyl software, in the Topomer CoMFA module, to training set S
1Compound divide substituting group, substituting group is divided on the one hand will guarantee that model meets theory, on the other hand the goodness of fit of model had certain influence, predictive ability to model is also very relevant simultaneously, and when connecting bridge only has a few structure, it as a substituting group, is conducive to inquire into connecting bridge to bioactive impact, so with the training set compound S
1Be divided into two substituting groups of connecting bridge and side chain, and adopt superimposed these 35 compounds of Topomer method; Molecular field with around the superimposed training set compound of probe scanning calculates MSA, CoMFA or CoMSIA molecular field, after molecular field information is selected, then with Y
1=LgBio is appointed as dependent variable and sets up linear model (called after M
1), institute's established model is the prime model of fit.Calculate the local physiological action P that compound substituent produces by Sybyl software in the modeling process
1Because compound has two substituting groups, so P
1Be bivector, in the molecule list, be expressed as Act_R
1And Act_R
2
3) structure of rear class model of fit
In SPSS Clementine software, with the above-mentioned local physiological action P that is calculated by Sybyl software
1Make independent variable, Y
1=LgBio makes dependent variable, carries out normalization with the activity of training set compound, obtains normalized value, goes normalization by neural network model, calculation training collection S
1The whole biologically active of compound is set up " thoroughly pruning " neural network model (called after M
2), institute's established model is the rear class model of fit, and sample is set to 100% to improve the predictive ability of model in the modeling process, and random seed is set to 0 to guarantee the repeatability of experiment.
Embodiment 2:
Present embodiment is that the goodness of fit is measured, and compares the M that above-described embodiment 1 is built
1-M
2Two-level model and M
1The goodness of fit of single-stage model, concrete steps are as follows:
1) variable naming
With model M
1To training set S
1The calculated activity called after Y of compound
2
With model M
2To training set S
1The calculated activity called after Y of compound
3
2) derive electronic form file
With Sybyl molecule list S
1.tbl the LgBio in and Pre_LgBio two row export as S
1_ M
1.csv file is converted to S again
1_ M
1.xls file.Above-mentioned LgBio is Y
1, Pre_LgBio is Y
2
Adopt identical method, from SPSS Clementine software, derive M2 to the training set compound S
1Calculated activity, save as S
1_ M
2.xls file; Wherein, S
1_ M
2.xls file including variable Y
1And Y
3
3) calculate related coefficient square and draw scatter diagram
With electrical form S
1_ M
1.xls file imports in the Origin software, to variable Y
1And Y
2Do linear regression, calculate related coefficient square R
1Be 0.95.Draw scatter diagram, the result as shown in Figure 1.
With electrical form S
1_ M
2.xls file imports in the Origin software, to variable Y
1And Y
3Do linear regression, calculate related coefficient square R
2Be 0.96.Draw scatter diagram, the result as shown in Figure 2.
Thus, can see employing front and back stages model M
1-M
2Compare the single-stage model M
1, related coefficient square R
2R
1=0.95, thus good capability of fitting had.
Embodiment 3:
Present embodiment is that estimated performance is measured, and compares the M that above-described embodiment 1 is built
1-M
2Two-level model and M
1The estimated performance of single-stage model, concrete steps are as follows:
1) arrangement of p38 kinase inhibiting activity
Get 35 non-training set S
1The pyrazole compound of element is set up test set S
2, its p38 kinase inhibiting activity is designated as Y
4With test set S
235 pyrazole compounds be made into Sybyl molecule list S
2.tbl, with Y
4Be appointed as dependent variable (in S2.tbl molecule list, being expressed as LgBio).
2) estimated performance of single-stage model M1 is measured
In the TopomerCoMFA module with Sybyl software, predictive molecule list S
2.tbl p38 kinase inhibiting activity, the result is designated as Y
5(at S
2.tbl be expressed as Pre_LgBio in the molecule list).In the forecasting process, calculate two substituent local physiological action P of compound
2, at S
2.tbl be expressed as Act_R in the molecule list
1And Act_R
2
In Origin software, calculate Y
4With Y
5Related coefficient square R
3Be 0.95, the scatter diagram that drafting obtains as shown in Figure 3.
3) estimated performance of two-level model M1-M2 is measured
In SPSS Clementine, with P
2Be independent variable, Y
4Be dependent variable, use the p38 kinase inhibiting activity of rear class model M 2 prediction test set compound Ss 2, the result is designated as Y
6
In Origin software, calculate Y
4With Y
6Related coefficient square R4 be 0.96, draw the scatter diagram obtain as shown in Figure 4.
Thus, can see employing front and back stages model M
1-M
2Compare the single-stage model M
1, related coefficient square R
4R
3=0.95, thus good estimated performance had.
The above; it only is the preferred embodiment of the invention; but protection scope of the present invention is not limited to this; anyly be familiar with those skilled in the art in scope disclosed in this invention; be equal to replacement or change according to technical scheme of the present invention and inventive concept thereof, all belonged to protection scope of the present invention.
Claims (10)
1. be used for the construction method of the two-stage match QSAR model of predictive compound activity, it is characterized in that may further comprise the steps:
1) get several compounds with identical skeleton as training set, the training set compound is divided substituting group, and superimposed training set compound;
2) according to the structure and activity of training set compound, adopt linear regression method to calculate the local physiological action that each substituting group produces, set up the prime model of fit;
3) according to activity and the step 2 of training set compound) the local physiological action that calculates, adopt neural network to calculate the whole biologically active of compound, set up the rear class model of fit;
4) with prime model of fit and the combination of rear class model of fit, be built into front and back stages QSAR model.
2. the construction method of the two-stage match QSAR model for the predictive compound activity according to claim 1 is characterized in that: step 2) activity of described training set compound is inhibition concentration or inhibiting rate.
3. the construction method of the two-stage match QSAR model for the predictive compound activity according to claim 2, it is characterized in that: step 1) is specific as follows:
For existing compound, carry out the data acquisition of biologic activity for the particular test system, data target adopts the negative logarithmic form of inhibition concentration or inhibiting rate, with this as the training set sample; Use the two-dimensional structure of Sybyl analysis software check compound, the compound by check is generated its three-dimensional structure; Subsequently, the substituting group of Further Division compound, and be optimized; At last, divide based on substituting group, and adopt the Topomer composite technology to carry out superimposed to above compound three-dimensional structure.
4. the construction method of each described two-stage match QSAR model for the predictive compound activity according to claim 3 is characterized in that: step 2) specific as follows:
With the molecular field around the superimposed training set compound of probe scanning, calculate MSA, CoMFA or CoMSIA molecular field, after molecular field information selected, carry out linear regression with the experiment activity of training set compound, obtain the prime model of fit of structure-activity relationship.
5. the construction method of each described two-stage match QSAR model for the predictive compound activity according to claim 4, it is characterized in that: step 3) is specific as follows:
With step 2) the local physiological action that calculates, carry out normalization with the activity of training set compound, obtain normalized value, go normalization by neural network model, calculate the whole biologically active of compound, obtain the rear class model of fit.
6. the construction method of each described two-stage match QSAR model for the predictive compound activity according to claim 1-5, it is characterized in that: described training set compound is the pyrazole compound with p38 kinase inhibition rate.
7. the construction method of each described two-stage match QSAR model for the predictive compound activity according to claim 1-5, it is characterized in that: the sample size of described training set compound has 30 at least.
8. the construction method of each described two-stage match QSAR model for the predictive compound activity according to claim 1-5, it is characterized in that: the substituting group that described training set compound is divided has 2 at least, includes the connecting bridge of compound in the substituting group of described division.
9. the construction method of each described two-stage match QSAR model for the predictive compound activity according to claim 1-5, it is characterized in that: the linear regression method that described step 2) adopts is partial least square method or principal component analysis (PCA).
10. the construction method of each described two-stage match QSAR model for the predictive compound activity according to claim 1-5, it is characterized in that: the neural network that described step 3) adopts is BF neural network or RBF neural network.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210455239.9A CN102930113B (en) | 2012-11-14 | 2012-11-14 | Building method of two-level fitting quantitative structure-activity relationship (QSAR) model for forecasting compound activity |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210455239.9A CN102930113B (en) | 2012-11-14 | 2012-11-14 | Building method of two-level fitting quantitative structure-activity relationship (QSAR) model for forecasting compound activity |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102930113A true CN102930113A (en) | 2013-02-13 |
| CN102930113B CN102930113B (en) | 2015-06-17 |
Family
ID=47644910
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201210455239.9A Expired - Fee Related CN102930113B (en) | 2012-11-14 | 2012-11-14 | Building method of two-level fitting quantitative structure-activity relationship (QSAR) model for forecasting compound activity |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102930113B (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103646180A (en) * | 2013-12-19 | 2014-03-19 | 山东大学 | Method for forecasting acute toxicity of organic compounds by building quantitative structure-activity relationship model with quantum chemistry method |
| CN104636619A (en) * | 2015-02-10 | 2015-05-20 | 青岛农业大学 | Method for rapidly and virtually screening human small intestine absorbable drugs |
| CN104834831A (en) * | 2015-04-08 | 2015-08-12 | 北京工业大学 | Consistency model building method based on 3-dimensional quantitative structure-activity relationship model |
| CN104866710A (en) * | 2015-05-08 | 2015-08-26 | 西北师范大学 | Method for predicting inhibition concentration of cytochrome P450 enzyme CYP1A2 inhibitor by utilizing simplified partial least squares |
| CN105787297A (en) * | 2016-03-12 | 2016-07-20 | 云南圣清环境监测科技有限公司 | Microbial remediation system activity evaluating method |
| CN108416184A (en) * | 2017-02-09 | 2018-08-17 | 清华大学深圳研究生院 | The 3D methods of exhibiting and system of compound |
| CN109360610A (en) * | 2018-11-26 | 2019-02-19 | 西南石油大学 | A kind of chemical molecular toxicity prediction model algorithm based on fuzzy neural network |
| CN110809800A (en) * | 2017-06-30 | 2020-02-18 | 学校法人明治药科大学 | Prediction device, prediction method, prediction program, learning model input data generation device, and learning model input data generation program |
| WO2020230043A1 (en) * | 2019-05-15 | 2020-11-19 | International Business Machines Corporation | Feature vector feasibilty estimation |
| CN112102900A (en) * | 2020-10-12 | 2020-12-18 | 北京晶派科技有限公司 | Drug design method based on TopoMA quantitative structure-activity relationship model |
| CN112151111A (en) * | 2020-08-27 | 2020-12-29 | 上海大学 | QSAR method for rapidly predicting xanthine derivative inhibitory activity based on multiple linear regression |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2000079263A2 (en) * | 1999-06-18 | 2000-12-28 | Synt:Em S.A. | Identifying active molecules using physico-chemical parameters |
-
2012
- 2012-11-14 CN CN201210455239.9A patent/CN102930113B/en not_active Expired - Fee Related
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2000079263A2 (en) * | 1999-06-18 | 2000-12-28 | Synt:Em S.A. | Identifying active molecules using physico-chemical parameters |
Non-Patent Citations (4)
| Title |
|---|
| 周喜斌等: "几种QSAR建模方法在化学中的应用与研究进展", 《计算机与应用化学》, vol. 28, no. 6, 28 June 2011 (2011-06-28), pages 761 - 764 * |
| 王威等: "基于交叉验证的逐步回归结合RBF神经网络在QSAR中的应用", 《南京化工大学学报》, vol. 23, no. 5, 30 October 2001 (2001-10-30) * |
| 肖方竹: "基于主成分分析_人工神经网络对氯苯酚毒性QSAR研究", 《毒理学杂志》, vol. 26, no. 5, 25 October 2012 (2012-10-25), pages 336 - 340 * |
| 闫宁: "HIV_1逆转录酶抑制剂的定量构效关系及分子设计研究", 《中国优秀硕士学位论文全文数据库医药卫生科技辑》, 15 January 2012 (2012-01-15) * |
Cited By (21)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103646180A (en) * | 2013-12-19 | 2014-03-19 | 山东大学 | Method for forecasting acute toxicity of organic compounds by building quantitative structure-activity relationship model with quantum chemistry method |
| CN104636619A (en) * | 2015-02-10 | 2015-05-20 | 青岛农业大学 | Method for rapidly and virtually screening human small intestine absorbable drugs |
| CN104636619B (en) * | 2015-02-10 | 2017-11-14 | 青岛农业大学 | A kind of method that quick virtual screening human small intestine easily absorbs the drug |
| CN104834831A (en) * | 2015-04-08 | 2015-08-12 | 北京工业大学 | Consistency model building method based on 3-dimensional quantitative structure-activity relationship model |
| CN104834831B (en) * | 2015-04-08 | 2017-06-16 | 北京工业大学 | A kind of consistency model construction method based on three-dimensional quantitative structure-activity relationship model |
| CN104866710A (en) * | 2015-05-08 | 2015-08-26 | 西北师范大学 | Method for predicting inhibition concentration of cytochrome P450 enzyme CYP1A2 inhibitor by utilizing simplified partial least squares |
| CN104866710B (en) * | 2015-05-08 | 2017-11-10 | 西北师范大学 | The method for predicting Cytochrome P450 1A2 inhibitor inhibition concentrations |
| CN105787297A (en) * | 2016-03-12 | 2016-07-20 | 云南圣清环境监测科技有限公司 | Microbial remediation system activity evaluating method |
| CN108416184B (en) * | 2017-02-09 | 2020-06-16 | 清华大学深圳研究生院 | 3D display method and system of compound |
| CN108416184A (en) * | 2017-02-09 | 2018-08-17 | 清华大学深圳研究生院 | The 3D methods of exhibiting and system of compound |
| CN110809800B (en) * | 2017-06-30 | 2024-01-02 | 学校法人明治药科大学 | Prediction device, prediction method, prediction program, learning model input data generation device, and learning model input data generation program |
| CN110809800A (en) * | 2017-06-30 | 2020-02-18 | 学校法人明治药科大学 | Prediction device, prediction method, prediction program, learning model input data generation device, and learning model input data generation program |
| CN109360610B (en) * | 2018-11-26 | 2019-11-15 | 西南石油大学 | A method of the chemical molecular toxicity prediction model based on fuzzy neural network |
| CN109360610A (en) * | 2018-11-26 | 2019-02-19 | 西南石油大学 | A kind of chemical molecular toxicity prediction model algorithm based on fuzzy neural network |
| WO2020230043A1 (en) * | 2019-05-15 | 2020-11-19 | International Business Machines Corporation | Feature vector feasibilty estimation |
| CN113795889A (en) * | 2019-05-15 | 2021-12-14 | 国际商业机器公司 | Feature vector feasibility estimation |
| GB2599520A (en) * | 2019-05-15 | 2022-04-06 | Ibm | Feature vector feasibilty estimation |
| US11798655B2 (en) | 2019-05-15 | 2023-10-24 | International Business Machines Corporation | Feature vector feasibility estimation |
| CN112151111A (en) * | 2020-08-27 | 2020-12-29 | 上海大学 | QSAR method for rapidly predicting xanthine derivative inhibitory activity based on multiple linear regression |
| CN112102900A (en) * | 2020-10-12 | 2020-12-18 | 北京晶派科技有限公司 | Drug design method based on TopoMA quantitative structure-activity relationship model |
| CN112102900B (en) * | 2020-10-12 | 2024-02-23 | 北京晶泰科技有限公司 | Drug design method based on TopoMA quantitative structure-activity relationship model |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102930113B (en) | 2015-06-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102930113B (en) | Building method of two-level fitting quantitative structure-activity relationship (QSAR) model for forecasting compound activity | |
| Le Bagousse-Pinguet et al. | Phylogenetic, functional, and taxonomic richness have both positive and negative effects on ecosystem multifunctionality | |
| May et al. | Moving beyond abundance distributions: neutral theory and spatial patterns in a tropical forest | |
| Dai et al. | Improving the Bi-LSTM model with XGBoost and attention mechanism: A combined approach for short-term power load prediction | |
| Dumitru et al. | Solar photovoltaic energy production forecast using neural networks | |
| Langmead et al. | Coral reef community dynamics and disturbance: a simulation model | |
| Acosta-Michlik et al. | Integrated assessment of sustainability trade-offs and pathways for global bioenergy production: Framing a novel hybrid approach | |
| CN102968672B (en) | Intelligent city based on housing preference plan model dynamic microscopic simulation method | |
| Wu | Optimization of AI-driven communication systems for green hospitals in sustainable cities | |
| Liu et al. | Application of the Fuzzy Neural Network Algorithm in the Exploration of the Agricultural Products E-Commerce Path. | |
| Zhang et al. | HOMER-based multi-scenario collaborative planning for grid-connected PV-storage microgrids with electric vehicles | |
| Li et al. | Prediction of grain yield in Henan Province based on grey BP neural network model | |
| CN106488482B (en) | Wireless sensor network optimizing method based on multi-Agent evolutionary Algorithm | |
| Hu et al. | Scenario reduction based on correlation sensitivity and its application in microgrid optimization | |
| Zhang | Research on safety simulation model and algorithm of dynamic system based on artificial neural network | |
| CN105279388A (en) | Multilayer cloud computing framework coordinated integrated reduction method for gestational-age newborn brain medical records | |
| Lu et al. | Application of GA optimized wavelet neural networks for carrying capacity of water resources prediction | |
| Uttej et al. | Prominent Technique for Rainfall Prediction using CatBoost over Light GBM for improving the Accuracy of Prediction | |
| Jin | Feasibility Analysis and Countermeasures of Psychological Health Training Methods for Volleyball Players Based on Artificial Intelligence Technology | |
| CN101464980A (en) | Artificial system and implementing method for public article investment experiment | |
| Song et al. | Reconceptualizing beta diversity: a hypervolume geometric approach | |
| Luo et al. | Study on deep learning political culture communication system in Universities under the perspective of postmodern media | |
| De Aguiar et al. | Using reservoir computing for forecasting of wind power generated by a wind farm | |
| Zhang | RETRACTED: Integrating the Big Data in Sports and Resource Interaction Using Artificial Neural Network | |
| Jiao et al. | Urban Green Space Planning and Design for Sponge City |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20150617 Termination date: 20201114 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |