Essentiality and Transcriptome-Enriched Pathway Scores Predict Drug-Combination Synergy

Area of Science:

• Computational Biology
• Systems Pharmacology
• Cancer Therapy

Background:

• Current computational models for drug synergy lack mechanistic interpretation and gene essentiality integration.
• Systems pharmacology models offer expanded scope for drug combination screening.

Purpose of the Study:

• Investigate drug combination synergy for cancer therapies using NCI ALMANAC data.
• Evaluate the predictive power of gene and pathway features in drug interactions.

Main Methods:

• Utilized logistic regression to analyze gene and pathway features for drug synergy prediction.
• Trained predictive models on NCI-60 cell lines, KEGG pathways, and drug pairs.
• Performed genome-wide association analyses to assess gene and pathway significance.

Main Results:

• Pathway features showed a stronger correlation with drug-combination synergy than gene features.
• Significant associations were found for both genes and pathways, but with little overlap in expression and essentiality.
• Validated four drug-combination pathways, including PI3K-AKT and AMPK signaling pathways, in 16 cell lines.

Conclusions:

• Pathways significantly outperform genes in predicting drug-combination synergy.
• Integrating gene expression and essentiality, considering their distinct mechanisms, is crucial for improving predictive accuracy.