CN116417093A - Drug target interaction prediction method combining transducer and graph neural network - Google Patents
Drug target interaction prediction method combining transducer and graph neural network Download PDFInfo
- Publication number
- CN116417093A CN116417093A CN202211552664.XA CN202211552664A CN116417093A CN 116417093 A CN116417093 A CN 116417093A CN 202211552664 A CN202211552664 A CN 202211552664A CN 116417093 A CN116417093 A CN 116417093A
- Authority
- CN
- China
- Prior art keywords
- drug
- neural network
- protein
- transducer
- graph
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 41
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 39
- 239000003596 drug target Substances 0.000 title claims description 23
- 230000003993 interaction Effects 0.000 title claims description 23
- 239000003814 drug Substances 0.000 claims abstract description 66
- 229940079593 drug Drugs 0.000 claims abstract description 66
- 102000004169 proteins and genes Human genes 0.000 claims abstract description 66
- 108090000623 proteins and genes Proteins 0.000 claims abstract description 66
- 150000001875 compounds Chemical class 0.000 claims abstract description 23
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 23
- 125000003275 alpha amino acid group Chemical group 0.000 claims abstract description 12
- 230000007246 mechanism Effects 0.000 claims abstract description 9
- 239000013598 vector Substances 0.000 claims description 17
- 238000000605 extraction Methods 0.000 claims description 12
- 230000008569 process Effects 0.000 claims description 8
- 238000012549 training Methods 0.000 claims description 8
- 125000000539 amino acid group Chemical group 0.000 claims description 3
- 238000012514 protein characterization Methods 0.000 claims description 3
- 238000012360 testing method Methods 0.000 claims description 3
- 230000002687 intercalation Effects 0.000 claims description 2
- 238000009830 intercalation Methods 0.000 claims description 2
- 238000013507 mapping Methods 0.000 claims description 2
- 238000010200 validation analysis Methods 0.000 claims description 2
- 238000010586 diagram Methods 0.000 abstract description 7
- 238000013136 deep learning model Methods 0.000 abstract description 3
- 230000000694 effects Effects 0.000 abstract description 3
- 238000012512 characterization method Methods 0.000 description 4
- 238000007876 drug discovery Methods 0.000 description 4
- 230000004913 activation Effects 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 239000004973 liquid crystal related substance Substances 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 239000002547 new drug Substances 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 239000000126 substance Chemical group 0.000 description 2
- 241000711573 Coronaviridae Species 0.000 description 1
- DFVKOWFGNASVPK-BWHPXCRDSA-N [cyano-(4-phenoxyphenyl)methyl] (1s,3s)-3-[(z)-2-chloro-3,3,3-trifluoroprop-1-enyl]-2,2-dimethylcyclopropane-1-carboxylate Chemical compound CC1(C)[C@H](\C=C(/Cl)C(F)(F)F)[C@@H]1C(=O)OC(C#N)C(C=C1)=CC=C1OC1=CC=CC=C1 DFVKOWFGNASVPK-BWHPXCRDSA-N 0.000 description 1
- 150000001413 amino acids Chemical class 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 230000009897 systematic effect Effects 0.000 description 1
- 238000012795 verification Methods 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
- G16C20/30—Prediction of properties of chemical compounds, compositions or mixtures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computing Systems (AREA)
- Data Mining & Analysis (AREA)
- Software Systems (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Mathematical Physics (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Crystallography & Structural Chemistry (AREA)
- Chemical & Material Sciences (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention represents the drug compound as a molecular diagram and the target protein as an amino acid sequence. In addition, the characteristic representations of the drug atoms and protein sequences are learned using a graph convolution neural network and a convolutional neural network, respectively. The obtained drug map information and protein local residue information are then input into a transducer encoder, respectively, so that the advantages of the map neural network and the transducer can be combined. Due to the multi-headed attentive mechanism in the transducer, the method can mine the atomic information of the drug molecules and the residue information in the protein sequence more deeply. The deep learning model is easy to be fitted when learning on small sample data, and the model is classified by combining a transducer and a graph neural network, so that the generalization capability of the model is improved, and a good effect is achieved on the small sample data. The method used by the invention is an end-to-end model, and compared with other prior art, the model can obviously improve the prediction precision.
Description
Technical Field
The invention belongs to the field of artificial intelligence, and particularly relates to a drug target interaction prediction method combining a transducer and a graph neural network.
Background
The prediction of drug-target interactions (DTI) plays a crucial role in drug discovery. The discovery and development of new drugs is both expensive and time consuming. A typical discovery process for new drugs typically costs between 5 and 26 billion dollars, taking 15 years. With the emergence of new diseases such as novel coronaviruses in 2019, drug discovery is more important. The recognition of interactions between drug compounds and target proteins plays a vital role in drug discovery and drug reuse. The drug discovery time can be shortened and the research cost can be effectively reduced by predicting drug-target interaction through a calculation method. Therefore, it is also highly desirable to develop accurate and efficient methods of predicting DTI. In recent years, the deep learning method has been widely used in DTI prediction. However, most of the studies in the prior art do not make full use of the molecular structure of the pharmaceutical compounds and the sequence structure of the proteins, so that these models cannot obtain accurate and efficient characterization.
Currently, the most widely used method for predicting drug-target interactions is to use different descriptors to represent drugs and proteins, respectively, and then input these representation vectors into various deep neural networks, such as convolutional neural networks, recurrent neural networks, and graph neural networks, to predict interactions. Numerous studies have shown that a model based on a graph neural network can effectively extract topology information from a drug molecular structure diagram. Furthermore, a transducer-based model has also been shown to be capable of extracting partial characteristic information of protein sequences and drug molecules. However, current drug-target interaction prediction studies based on graph neural networks mostly focus on capturing the node features of drug molecules, while the extraction and processing of edge features is omitted. This may lead to loss of important structural information after construction of the drug molecular diagram. On the other hand, many methods represent proteins as amino acid sequences and encode them using convolutional neural networks. However, existing shallow convolutions are insufficient to capture the structural features of the primary structure of the protein.
Disclosure of Invention
The invention aims at: a method for predicting drug-target interactions by combining a transducer and a graphic neural network is provided, wherein the protein and the drug are represented by using an amino acid sequence and a molecular diagram respectively, and are encoded by using the convolutional neural network and the graphic neural network respectively, and the final characteristic representation of data obtained by the transducer is used for predicting the interactions of given drug-target pairs.
The technical scheme of the invention is as follows: a method for predicting drug target interactions in combination with a transducer and a graph neural network, comprising the steps of:
(1) Data represent, drug compounds in the dataset as molecular figures, target proteins as amino acid sequences;
(2) Feature extraction, namely learning feature representations of the drug compounds and the protein sequences by using a graph convolution neural network and a convolution neural network respectively, and independently inputting the obtained drug graph information and protein amino acid residue information into a transducer encoder to obtain feature representations of the drug compounds and the protein;
(3) A prediction module links together the feature representations extracted from the drug compounds and proteins to obtain a combined representation of a given drug target pair, the feature vectors of which are input into the multi-layer perceptron to obtain a final DTI prediction.
Further, the step (1) specifically includes selecting a data set and dividing the data set into a training set, a verification set and a test set; the drug compound is input in the form of SMILES codes, and the SMILES codes of the input drug compound are converted into a graph structure through RDkit in the pretreatment process; proteins are input in the form of amino acid sequences, for which an intercalation procedure is used prior to training.
Further, the selected data sets are Human and c.elegans data sets.
Further, the step (2) is specifically that,
(a) Using a three-layer graph convolution neural network, obtaining a graph-level characteristic representation of each drug compound through the network, and mapping an input graph into a graph-level representation vector by a multi-layer graph convolution layer through a two-stage process;
(b) Adopting a convolutional neural network architecture of three different layers to detect local residue information of proteins, and executing a global maximum pool to capture important local residual characteristics after the different layers of the convolutional neural network; these three eigenvectors obtained from three different layers of the convolutional neural network are then concatenated and input into the linear layer to obtain a protein characterization;
(c) The two used transducer encoders calculate the context representation of the drug features and the protein features in parallel, so that the multi-head attention mechanism can learn the similarity and the dependence among the elements and complete the feature extraction.
Further, step (3) is specifically to link the two characteristic representations of the drug and protein, respectively, with the previously obtained graphic neural network drug characteristics and convolutional neural network protein characteristics, after they are obtained by a transducer encoder, and then link them together and input into a full link layer module to predict the likelihood of a given drug target interaction.
The invention has the advantages that:
the invention provides an end-to-end deep learning method combining a transducer and a graph neural network. In this model, the drug compounds are represented as molecular figures and the target proteins are represented as amino acid sequences. In addition, the characteristic representations of drug atoms and protein sequences are also learned using a graph convolutional neural network and a convolutional neural network, respectively. The obtained drug map information and protein local residue information are then input into a transducer encoder, respectively, so that the advantages of the map neural network and the transducer can be combined. Due to the multi-headed attentive mechanism in the transducer, the method of the present invention can mine the atomic information of the drug molecule and the residue information in the protein sequence more deeply. The deep learning model is easy to be fitted when learning on small sample data, and the model is classified by combining a transducer and a graph neural network, so that the generalization capability of the model is improved, and a good effect is achieved on the small sample data. The method used by the invention is an end-to-end model, and compared with other prior art, the model can obviously improve the prediction precision.
Drawings
The invention is further described below with reference to the accompanying drawings and examples:
FIG. 1 is a flow chart of a method of drug target interaction prediction incorporating a transducer and a graph neural network of the present invention.
Fig. 2 is a feature extraction block diagram of drug data.
FIG. 3 is a block diagram of feature extraction of protein data.
Detailed Description
Examples: the method for predicting the drug-target interaction by combining a transducer and a graphic neural network comprises the following steps of:
1. the data represents: drug compounds in the dataset are represented as molecular figures, and target proteins are represented as amino acid sequences;
2. feature extraction: the characteristic representations of the drug atoms and protein sequences are learned using a graph convolution neural network and a convolutional neural network, respectively. The obtained drug profile information and protein amino acid residue information are then independently input into a transducer encoder, whereby we can combine the advantages of the profile neural network and the transducer. Our invention can mine the atomic information of drug molecules and the residue information in protein sequences deeper due to the multi-headed attentive mechanism in the transducer.
3. And a prediction module: in this module, the feature representations extracted from the drug and protein are linked together to obtain a combined representation of a given drug-target pair. The combined feature vectors are input into a multi-layer perceptron to obtain the final DTI prediction.
As shown in fig. 1. The method comprises the following specific steps:
1. the data represents:
to obtain the drug and protein data required for the experiment, the method selects two commonly used reference data sets, namely a Human and a c.elegans data set, which are widely used for binary classification tasks. The Human and c.elegans datasets contained positive drug-target pair samples from different databases, including DrugBank, matador and STITCH databases, as well as negative samples obtained by the systematic screening method proposed by Liu et al. From previous correlation studies, we randomly split the two data sets into training, validation and test sets at a ratio of 8:1:1 to compare our model with other methods. The following table one summarizes the information of the two reference data sets:
detailed information of table one two data sets
Data set | Quantity of drug | Protein quantity | Number of interactions |
Human | 2726 | 2001 | 6726 |
C.elegans | 1767 | 1876 | 7786 |
Drugs and proteins first need to be processed and presented before they can be delivered to the neural network. In the present invention, the drug is administered in the form of a SMILES code, which describes the three-dimensional structure of the compound by means of a string. On the other hand, proteins are input in the form of amino acid sequences. Then, during the pretreatment process, the input compound SMILES is converted into a graph structure by RDkit, wherein nodes and edges represent atoms and chemical bonds between atoms, respectively. Wherein, the liquid crystal display device comprises a liquid crystal display device, each atomic feature is represented as a vector of length 41. For protein amino acid sequences, an embedding procedure was used prior to training. Specifically, we extract unique letters representing the class of amino acids, each letter further represented by an integer. These integers can then be used to convert the protein sequence into a code. To improve the convenience of the training process, we set the maximum length of the protein sequence to 1000 according to previous studies, thereby covering most of the proteins in the dataset.
2. Feature extraction:
2.1 neural network Module
Further, we rolled up a neural network using a three-layer graph, through which a graphical level characterization of each drug was obtained. In general, the graph is represented as g= (V, E), where V is the node set and E is the edge set. The multi-layer graph convolution layer maps the input graph into a graph-level representation vector through a two-stage process.
In the first phase, the messaging phase, features of the graph are extracted by implementing the following two steps: information is collected and updated. Each atom x i Collecting local information t from its neighboring atoms and edges i . Atomic node x is then updated using the graph volume stacking i The following is shown:
wherein W is 1 、W 2 Is a weight vector that can be learned and,for the node characteristic of node i at time t, < >>Is a neighbor of node i, σ represents the ReLU activation function. Processing the slave through the above formulaAtomic characteristic information learned in the previous step to obtain updated characteristics at each iteration.
Then, in a second phase, the read phase, node features for a given drug compound are globally aggregated. In the read phase, to obtain the final feature vector from the set of node vectors in G, we embed the nodes as follows:
wherein the method comprises the steps ofRepresenting the number of atoms in the drug molecule, L is the final iterative step. The read phase aggregates the node level representation into a 64-dimensional graph level feature representation to meet the needs of the present invention.
2.2 convolutional neural network module
In this module we constructed a multi-layer convolutional neural network to extract deep features of proteins. Specifically, we designed a convolutional neural network architecture with three different layers to detect local residue information of proteins. After convolving the different layers of the neural network, a global maximum pool is performed to capture important local residual features. Finally, we join the three eigenvectors obtained from three different layers of the convolutional neural network and then input them into the linear layer to obtain the protein characterization. In our invention we set the dimension of the final protein representation to 64, the same as the dimension of the drug representation.
2.3 Transformer module
To capture biological and chemical information of drugs and proteins, respectively, we used two parallel transducer encoders. Specifically, the output vectors of the graph convolutional neural network for drug feature extraction and the convolutional neural network for protein feature extraction are then input into two encoders.
A key part of the transducer encoder block is the multi-headed attention module. The multi-headed interest module includes a plurality of scaled point-attention layers to extract relationship information and calculate the similarity of each element in the input vector. The multi-headed attention layer contains three different vectors converted from the input, including key K, value V and query Q, which are generated from the input sequence by the linear layer and split between the different attention heads, and then calculate the attention score. Further, each header maps a query Q and a key-value pair to an output, the output is calculated from a weighted sum of values. The attention weight assigned to each value is obtained by applying SoftMax to the scaled dot product between the query and the key. The outputs of each attention head are connected together, the final dimension of the output vector being the same as the input vector. The formula for the multi-headed attention layer is as follows:
MultiHead(Q,K,V)=Concat(head 1 ,...,head n )W 0
wherein d is k The input size, Q, K, V, is the key, value and query obtained from the input sequence through the linear layer.W 0 As the weight matrix, n represents the number of heads of the multi-head attention mechanism, and is set to 4 in the present method.
Further, after the multi-head attention module, each encoder contains a full link layer and a dropout module, as follows:
FFN(x)=ReLU(xW 1 +b 1 )W 2 +b 2
wherein W is 1 、W 2 Representing a learnable weight matrix, b 1 、b 2 Is a bias term. The attention mechanism enables our approach to focus on some important and critical parts of the input data so that the model can directly capture the interaction and relationship information of the obtained drug and protein feature vectors, respectively. Furthermore, according to some previous studies, consideration was given to elemental sequence pairs of drug molecules and protein sequencesIn the prediction of drug-target interactions, we deleted the position coding in the transducer model. In general, two transducer encoders used in the present invention compute contextual representations of drug features and protein features in parallel, whereby a multi-headed attention mechanism is able to learn the similarity and dependence between individual elements.
The feature extraction modules of the drug and protein data are shown in figures 2 and 3, respectively, below.
3. And a prediction module:
after two 128-dimensional characterization of the drug and protein were obtained by the transducer encoder, we connected them to the previously obtained graph neural network drug and convolutional neural network protein features, respectively. In this way we have obtained two 192-dimensional characterization representations of drug and protein, respectively. They are then linked together and input into a full link layer module to predict the likelihood of a given drug-target interaction. In particular, the fully connected layer in the present invention comprises three layers to convert the obtained feature representation into interaction probabilities. Wherein each linear layer is arranged to follow the ReLU activation function and the dropout layer. Considering that DTI prediction in our experiments is a binary classification task, we use cross entropy loss for model training as follows:
wherein the method comprises the steps ofAnd y i The predicted probability and true probability for a given drug-target pair are represented, respectively, with n being the sample size.
The invention represents the drug compound as a molecular diagram and the target protein as an amino acid sequence. In addition, the characteristic representations of the drug atoms and protein sequences are learned using a graph convolution neural network and a convolutional neural network, respectively. The obtained drug map information and protein local residue information are then input into a transducer encoder, respectively, so that the advantages of the map neural network and the transducer can be combined. Due to the multi-headed attentive mechanism in the transducer, the method can mine the atomic information of the drug molecules and the residue information in the protein sequence more deeply. The deep learning model is easy to be fitted when learning on small sample data, and the model is classified by combining a transducer and a graph neural network, so that the generalization capability of the model is improved, and a good effect is achieved on the small sample data. The method used by the invention is an end-to-end model, and compared with other prior art, the model can obviously improve the prediction precision.
The above embodiments are merely for illustrating the technical concept and features of the present invention, and are not intended to limit the scope of the present invention to those skilled in the art to understand the present invention and implement the same. All modifications made according to the spirit of the main technical proposal of the invention should be covered in the protection scope of the invention.
Claims (5)
1. A method for predicting drug target interactions in combination with a transducer and a graph neural network, comprising the steps of:
(1) Data represent, drug compounds in the dataset as molecular figures, target proteins as amino acid sequences;
(2) Feature extraction, namely learning feature representations of the drug compounds and the protein sequences by using a graph convolution neural network and a convolution neural network respectively, and independently inputting the obtained drug graph information and protein amino acid residue information into a transducer encoder to obtain feature representations of the drug compounds and the protein;
(3) A prediction module links together the feature representations extracted from the drug compounds and proteins to obtain a combined representation of a given drug target pair, the feature vectors of which are input into the multi-layer perceptron to obtain a final DTI prediction.
2. The method of claim 1, wherein step (1) is specifically selecting a dataset and classifying the dataset into a training set, a validation set and a test set; the drug compound is input in the form of SMILES codes, and the SMILES codes of the input drug compound are converted into a graph structure through RDkit in the pretreatment process; proteins are input in the form of amino acid sequences, for which an intercalation procedure is used prior to training.
3. The method of claim 2, wherein the selected dataset is a Human and c.
4. The method for predicting drug target interactions binding to a transducer and a graphic neural network according to claim 2, wherein the step (2) is specifically,
(a) Using a three-layer graph convolution neural network, obtaining a graph-level characteristic representation of each drug compound through the network, and mapping an input graph into a graph-level representation vector by a multi-layer graph convolution layer through a two-stage process;
(b) Adopting a convolutional neural network architecture of three different layers to detect local residue information of proteins, and executing a global maximum pool to capture important local residual characteristics after the different layers of the convolutional neural network; these three eigenvectors obtained from three different layers of the convolutional neural network are then concatenated and input into the linear layer to obtain a protein characterization;
(c) The two used transducer encoders calculate the context representation of the drug features and the protein features in parallel, so that the multi-head attention mechanism can learn the similarity and the dependence among the elements and complete the feature extraction.
5. The method of claim 4, wherein the step (3) is specifically to connect the two characteristic representations of the drug and the protein with the previously obtained characteristic of the drug of the graphic neural network and the characteristic of the protein of the convolutional neural network, respectively, after obtaining the two characteristic representations of the drug and the protein by the transducer encoder, and then connect them together and input them into the full connection layer module to predict the possibility of the interaction of the given drug target.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211552664.XA CN116417093A (en) | 2022-12-06 | 2022-12-06 | Drug target interaction prediction method combining transducer and graph neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211552664.XA CN116417093A (en) | 2022-12-06 | 2022-12-06 | Drug target interaction prediction method combining transducer and graph neural network |
Publications (1)
Publication Number | Publication Date |
---|---|
CN116417093A true CN116417093A (en) | 2023-07-11 |
Family
ID=87057078
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202211552664.XA Pending CN116417093A (en) | 2022-12-06 | 2022-12-06 | Drug target interaction prediction method combining transducer and graph neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116417093A (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116486900A (en) * | 2023-04-25 | 2023-07-25 | 徐州医科大学 | Drug target affinity prediction method based on depth mode data fusion |
CN116992919A (en) * | 2023-09-28 | 2023-11-03 | 之江实验室 | Plant phenotype prediction method and device based on multiple groups of science |
CN117393036A (en) * | 2023-11-09 | 2024-01-12 | 中国海洋大学 | Protein multi-level semantic polymerization characterization method for drug-target affinity prediction |
CN117393050A (en) * | 2023-10-17 | 2024-01-12 | 哈尔滨工业大学(威海) | Protein function recognition method, device and readable storage medium |
CN117912591A (en) * | 2024-03-19 | 2024-04-19 | 鲁东大学 | Kinase-drug interaction prediction method based on deep contrast learning |
CN117912591B (en) * | 2024-03-19 | 2024-05-31 | 鲁东大学 | Kinase-drug interaction prediction method based on deep contrast learning |
-
2022
- 2022-12-06 CN CN202211552664.XA patent/CN116417093A/en active Pending
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116486900A (en) * | 2023-04-25 | 2023-07-25 | 徐州医科大学 | Drug target affinity prediction method based on depth mode data fusion |
CN116486900B (en) * | 2023-04-25 | 2024-05-03 | 徐州医科大学 | Drug target affinity prediction method based on depth mode data fusion |
CN116992919A (en) * | 2023-09-28 | 2023-11-03 | 之江实验室 | Plant phenotype prediction method and device based on multiple groups of science |
CN116992919B (en) * | 2023-09-28 | 2023-12-19 | 之江实验室 | Plant phenotype prediction method and device based on multiple groups of science |
CN117393050A (en) * | 2023-10-17 | 2024-01-12 | 哈尔滨工业大学(威海) | Protein function recognition method, device and readable storage medium |
CN117393036A (en) * | 2023-11-09 | 2024-01-12 | 中国海洋大学 | Protein multi-level semantic polymerization characterization method for drug-target affinity prediction |
CN117912591A (en) * | 2024-03-19 | 2024-04-19 | 鲁东大学 | Kinase-drug interaction prediction method based on deep contrast learning |
CN117912591B (en) * | 2024-03-19 | 2024-05-31 | 鲁东大学 | Kinase-drug interaction prediction method based on deep contrast learning |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN113707235B (en) | Drug micromolecule property prediction method, device and equipment based on self-supervision learning | |
Tan et al. | Instance-level image retrieval using reranking transformers | |
CN116417093A (en) | Drug target interaction prediction method combining transducer and graph neural network | |
CN110289050B (en) | Drug-target interaction prediction method based on graph convolution sum and word vector | |
CN112199532B (en) | Zero sample image retrieval method and device based on Hash coding and graph attention machine mechanism | |
CN111400494B (en) | Emotion analysis method based on GCN-Attention | |
Zhang et al. | Learning implicit class knowledge for RGB-D co-salient object detection with transformers | |
CN110188827A (en) | A kind of scene recognition method based on convolutional neural networks and recurrence autocoder model | |
CN113673346A (en) | Motor vibration data processing and state recognition method based on multi-scale SE-Resnet | |
CN116072227B (en) | Marine nutrient biosynthesis pathway excavation method, apparatus, device and medium | |
CN111008224A (en) | Time sequence classification and retrieval method based on deep multitask representation learning | |
CN114913938B (en) | Small molecule generation method, equipment and medium based on pharmacophore model | |
CN115985386A (en) | Multi-modal drug-protein target interaction prediction method and system | |
CN116758397A (en) | Single-mode induced multi-mode pre-training method and system based on deep learning | |
CN117237559A (en) | Digital twin city-oriented three-dimensional model data intelligent analysis method and system | |
CN109857892B (en) | Semi-supervised cross-modal Hash retrieval method based on class label transfer | |
CN114579794A (en) | Multi-scale fusion landmark image retrieval method and system based on feature consistency suggestion | |
CN110516640B (en) | Vehicle re-identification method based on feature pyramid joint representation | |
CN112085245A (en) | Protein residue contact prediction method based on deep residual error neural network | |
CN112116949A (en) | Protein folding identification method based on triple loss | |
CN116955650A (en) | Information retrieval optimization method and system based on small sample knowledge graph completion | |
CN115458044A (en) | Medicine and medicine interaction prediction method based on biological network global structure | |
CN111259176B (en) | Cross-modal Hash retrieval method based on matrix decomposition and integrated with supervision information | |
CN112735604B (en) | Novel coronavirus classification method based on deep learning algorithm | |
CN110795591B (en) | Image retrieval method based on discrete gradient back propagation |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |