Multi-view graph convolutional network for cancer cell-specific synthetic lethality prediction.

Area of Science:

• Computational biology
• Genomics
• Bioinformatics

Background:

• Synthetic lethal (SL) interactions are crucial for developing targeted cancer therapies.
• Wet-lab screening for SL gene pairs is time-consuming and expensive.
• Existing computational methods, including graph neural networks (GNNs), often lack cancer cell specificity and fail to fully leverage diverse biological features.

Purpose of the Study:

• To develop a novel computational method for predicting cancer cell-specific SL gene pairs.
• To improve the accuracy and efficiency of identifying potential therapeutic targets for cancer.

Main Methods:

• Proposed MVGCN-iSL, a multi-view graph convolutional network (GCN) model.
• Integrated five biological graph features and multi-omics data.
• Utilized max pooling for integrating GCN representations and a deep neural network (DNN) for prediction of cancer cell-specific SL interactions (iSL).

Main Results:

• MVGCN-iSL successfully integrated multiple graph features and multi-omics data.
• The model achieved state-of-the-art performance in predicting SL gene pairs.
• Demonstrated strong generalization ability to novel genes.

Conclusions:

• MVGCN-iSL offers a powerful computational approach for discovering cancer cell-specific SL interactions.
• The model enhances the potential for developing targeted cancer therapeutics.
• Highlights the importance of integrating diverse biological features for improved predictive accuracy.