Self-Attention Neural Network for Drug-Target Binding Region Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current drug-target interaction prediction methods using deep learning models, such as CNN and transformer-based models, do not explicitly train the binding region of a drug, limiting their accuracy in predicting drug-target interactions.
Innovation Solution
A self-attention-based deep neural network is used to predict the binding region and drug-target interaction by training a transformer network with drug fingerprint and protein sequence data, transforming the data into drug tokens and protein grid encodings, and inputting them to the network to predict the interaction between the drug and protein, incorporating binding information and three-dimensional structures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional deep learning models (CNN, RNN, transformer) are used for drug-target interaction prediction, then the model can process drug databases and make predictions, but the prediction accuracy is limited because the binding region is not explicitly trained
Solution Approach 1:
The patent segments the drug molecule into a binding region and non-binding region through explicit training. The model divides the drug representation into distinct functional parts, allowing the binding region to be identified and processed separately, which improves prediction accuracy by focusing computational attention on the most relevant molecular segments.
Solution Approach 2:
The patent introduces a new dimensional aspect to the prediction model by explicitly adding binding region prediction as a separate output dimension alongside DTI prediction. This multi-dimensional approach allows the model to simultaneously predict both the interaction outcome and the specific binding region, thereby improving overall prediction accuracy without merely expanding the model size.
2Measurement precision
If the binding region is explicitly trained in the deep learning model, then the accuracy of binding region prediction increases, but the device complexity and training difficulty increase
Solution Approach 1:
The patent implements a multi-functional deep learning model that simultaneously performs drug-target interaction prediction and binding region prediction. By designing the model to handle multiple prediction tasks within a single architecture, the system achieves binding region identification capability without requiring entirely separate models, thus managing complexity while improving accuracy.
Solution Approach 2:
The patent incorporates preliminary action by pre-processing drug molecular structures into representations that highlight potential binding regions before the main prediction process. This pre-processing step prepares the data in a format that facilitates easier identification of binding regions during training, reducing the complexity burden on the core prediction model.
Data Source
AI summary
The present invention relates to drug-target protein interaction prediction using deep learning, and a device and a method for predicting a drug-target interaction (DTI), according to the present invention, train a transformer network by using the interaction between a drug and a protein, and the binding region of the drug and the protein, and predict DTI and the binding region by using the transformer network using an attention score, and thus DTI prediction performance can be increased.