SynPathy: Predicting Drug Synergy through Drug-Associated Pathways Using Deep Learning.

Area of Science:

• Computational biology
• Bioinformatics
• Machine learning in drug discovery

Background:

• Drug combination therapy is a key strategy in cancer treatment.
• High-throughput screening for synergistic drug combinations is limited by the vast number of potential agents and cell lines.
• Predicting drug synergy requires understanding complex molecular interactions.

Purpose of the Study:

• To develop a biologically-motivated deep learning approach for predicting drug synergy.
• To identify pathway-level features from molecular data for synergy prediction and interaction quantification.
• To explore the relationship between pathway proximity and drug synergy.

Main Methods:

• Utilized deep learning to analyze drug and cell line molecular data.
• Identified pathway-level features to predict drug synergy.
• Quantified interactions in synergistic drug pairs.
• Analyzed pathway proximity on protein interaction networks.

Main Results:

• Achieved a Mean Squared Error (MSE) of 70.6 ± 6.4, outperforming previous methods.
• Identified potential candidate pathways that explain drug synergy predictions.
• Demonstrated that drug combinations with closer top contributing pathways on protein interaction networks exhibit higher synergy.

Conclusions:

• The deep learning approach effectively predicts drug synergy and quantifies interactions.
• Pathway proximity on protein interaction networks is a potential indicator for synergistic drug combinations.
• The computational framework can aid in prescreening and designing novel drug combinations for cancer therapy.