Modelling Drug-Target Binding Affinity using a BERT based Graph Neural network.

Area of Science:

• Computational Biology
• Drug Discovery
• Artificial Intelligence

Background:

• Identifying drug-protein interactions is crucial for disease research but is often time-consuming and expensive.
• Deep learning offers a promising alternative for simulating these interactions efficiently.
• Existing methods may not fully leverage the power of large pre-trained models for this task.

Purpose of the Study:

• To develop and evaluate a novel deep learning framework for predicting drug-target interactions.
• To improve the accuracy and efficiency of drug-target interaction prediction.
• To reduce the cost and time associated with discovering novel drug-protein interactions.

Main Methods:

• Utilized BERT-style models pre-trained on extensive drug and protein datasets.
• Employed graph convolutional neural networks (GCNs) with BERT encodings as node representations.
• Modeled drug-target interactions as a graph structure for prediction.
• Performed interpretability analyses to identify functionally relevant regions in drugs and proteins.

Main Results:

• Achieved significant improvements over baseline BERT and previous state-of-the-art methods on benchmark datasets.
• Demonstrated the effectiveness of combining pre-trained BERT models with GCNs for drug-target interaction prediction.
• Successfully identified functionally relevant areas within drugs and proteins through interpretability analyses.

Conclusions:

• The proposed approach offers a powerful and efficient method for predicting drug-target binding affinity.
• Leveraging large pre-trained models and graph neural networks advances the field of computational drug discovery.
• This method has the potential to accelerate the identification of novel therapeutics.