MVGAE: A Multi-View Graph Auto-Encoder Model for Drug Prediction of Non-Small Cell Lung Cancer Based on Synthetic...

Area of Science:

• Computational biology
• Genomics
• Drug discovery

Background:

• Non-small cell lung cancer (NSCLC) requires novel therapeutic targets and drugs.
• Existing methods for identifying therapeutic targets and drugs are limited.

Purpose of the Study:

• To develop an integrated computational framework for identifying therapeutic target genes and drugs for NSCLC.
• To leverage multi-source biological network data for enhanced data mining and prediction.

Main Methods:

• Utilized a multi-view graph auto-encoder (MVGAE) model integrated with network-informed adaptive positive-unlabeled (NIAPU) and synthetic lethality multi-view graph auto-encoder (SLMGAE) models.
• Integrated protein-protein interaction networks, disease-gene associations, and gene-drug bipartite graphs.
• Employed computational screening to predict driver genes, synthetic lethality targets, and candidate drugs.

Main Results:

• Identified seven potential driver genes associated with NSCLC using the NIAPU model.
• Predicted nine genes with synthetic lethality (SL) interactions as candidate therapeutic targets using SLMGAE.
• Predicted corresponding targeted drugs for SL targets using MVGAE, with some targets (e.g., ATR, RAD51) showing conditional SL effects and some drugs (e.g., PAZOPANIB) previously linked to NSCLC treatment.

Conclusions:

• The integrated MVGAE framework effectively prioritizes potential therapeutic targets and candidate drugs for NSCLC.
• This approach offers a robust computational basis for drug repurposing and precision treatment strategies in NSCLC.
• The study highlights the potential of integrating complementary models for advancing NSCLC therapy.