TranSynergy

TranSynergy predicts synergistic drug combinations and deconvolutes the biological pathways underlying drug synergy using a mechanism-driven deep learning framework for cancer drug discovery.

Key Features:

• Mechanism-Driven Modeling: Explicitly models cellular effects of drug actions via cell-line gene dependency, gene-gene interactions, and genome-wide drug-target interactions.
• Self-Attention Boosted Deep Learning: Employs a self-attention mechanism within a deep learning architecture to focus on features associated with combination synergy.
• Knowledge Integration: Integrates biological knowledge of cellular and molecular interactions into the predictive framework.
• Interpretability through SA-GSEA: Implements Shapley Additive Gene Set Enrichment Analysis (SA-GSEA) to deconvolute pathways contributing to predicted synergy.

Scientific Applications:

• Drug Combination Screening: Predicts and prioritizes synergistic drug combinations for experimental validation.
• Precision Medicine: Identifies pathways and potential biomarkers to inform personalized treatment strategies.
• Therapeutic Discovery: Facilitates discovery of novel synergistic anti-cancer combinations, including applications in ovarian cancer.

Methodology:

Mechanism-driven deep learning architecture augmented with self-attention; explicit modeling of cell-line gene dependency, gene-gene interactions, and genome-wide drug-target interactions; and pathway deconvolution using SA-GSEA (Shapley Additive Gene Set Enrichment Analysis).