RAREsim2: flexible simulation of rare variant genetic data using real haplotypes

Area of Science:

• Computational biology
• Pharmacogenomics
• Artificial intelligence in drug discovery

Background:

• Traditional drug synergy screening is inefficient and costly.
• Current deep learning models for drug synergy lack causal gene-response modeling.
• Understanding gene-drug response causality is key for effective combination therapies.

Purpose of the Study:

• To propose CADS, a deep learning framework integrating causal gene relationships for drug synergy prediction.
• To accurately predict drug synergy while discovering interpretable causal genes.
• To advance AI-driven drug development with enhanced biological interpretability.

Main Methods:

• Developed CADS (Causal Adjustment for Drug Synergy) framework using multi-omics data.
• Integrated causal gene-drug response relationships via a learnable mask mechanism.
• Employed backdoor adjustment to filter irrelevant genetic factors.

Main Results:

• CADS consistently outperformed state-of-the-art methods across multiple metrics.
• Achieved accurate drug synergy prediction and interpretable causal gene discovery.
• Case studies identified clinically validated cancer genes mediating drug interactions.

Conclusions:

• CADS advances combination therapy prediction by modeling drug synergy causal genes.
• The framework offers enhanced interpretability for AI-based drug development.
• CADS provides valuable biological insights through gene importance scores.