Identification of pathways associated with chemosensitivity through network embedding.

Area of Science:

• Pharmacogenomics
• Computational Biology
• Systems Biology

Background:

• Basal gene expression predicts drug response but doesn't capture complex genotype-phenotype links.
• Molecular networks and biological pathways offer deeper insights into drug response variations.

Purpose of the Study:

• To identify specific pathways mediating chemical response by integrating chemosensitivity and gene expression data.
• To develop a computational method (PACER) for ranking pathway enrichment using network embedding.

Main Methods:

• Developed PACER, a network embedding method to represent genes and pathways in a shared vector space.
• Ranked pathways based on their relevance to genes correlated with cytotoxic drug response.
• Integrated pharmacogenomics data with CCLE gene expression data and known molecular networks.

Main Results:

• PACER effectively identifies compound-pathway associations, outperforming existing statistical methods.
• Demonstrated advantages on benchmarks using compound target genes and drug response signatures.
• Identified testable hypotheses for chemosensitivity pathways and drug mechanisms of action.

Conclusions:

• PACER offers a novel technique for identifying enriched properties within gene sets using network information.
• This method enhances the study of drug mechanisms and personalized medicine.
• PACER advances the analysis of complex biological networks for drug discovery.