Integrating Graph Convolution and Attention Mechanism for Kinase Inhibition Prediction.

Area of Science:

• Biochemistry and Molecular Biology
• Computational Biology and Cheminformatics
• Pharmacology and Drug Discovery

Background:

• Kinases are crucial enzymes in cell signaling; their dysregulation is linked to various human cancers and diseases.
• Targeting aberrant kinases with small drug molecules is a key strategy in cancer therapy.
• Previous efforts utilized machine learning and deep learning for kinase inhibition prediction, but advancements are ongoing.

Purpose of the Study:

• To develop and evaluate a Graph Neural Network (GNN) model for predicting kinase inhibition activities.
• To compare the performance of a standalone Graph Convolution Network (GCN) with a combined GCN and Graph Attention Network (GCN-GAT).
• To assess the predictive accuracy of the developed models on independent kinase datasets.

Main Methods:

• Development of two GNN models: a GCN and a GCN-GAT.
• Training and 10-fold cross-validation on two large kinase datasets comprising small drug molecules and targeted kinases.
• Evaluation of model performance on independent datasets using metrics like accuracy, MCC, sensitivity, specificity, and precision.

Main Results:

• The combined GCN-GAT model demonstrated superior performance compared to previous methods on both independent kinase datasets.
• On independent Kinase Dataset 1, the GCN-GAT model achieved accuracy of 0.96, MCC of 0.89, sensitivity of 0.90, specificity of 0.98, and precision of 0.91.
• On independent Kinase Dataset 2, the GCN-GAT model achieved accuracy of 0.97, MCC of 0.90, sensitivity of 0.91, specificity of 0.99, and precision of 0.92.

Conclusions:

• The GCN-GAT model is highly effective for predicting kinase inhibition activities.
• This approach offers a promising computational tool for accelerating the discovery of novel kinase-targeted drugs.
• The findings highlight the potential of advanced graph neural networks in pharmaceutical research and development.