Association analysis of the perturbation of interactions in biological pathways and anticancer drug activity

Area of Science:

• Oncology
• Genomics
• Systems Biology

Background:

• Precision medicine in cancer requires understanding genomic alterations and drug responses.
• Current research often analyzes genomic and drug-response profiles separately.
• A network perspective, focusing on signaling pathway interactions, is underutilized.

Purpose of the Study:

• To develop and evaluate an interaction-centric approach for predicting anticancer drug response.
• To identify novel drug-response associations by analyzing perturbed signaling pathways.
• To compare the efficacy of interaction-centric versus gene-centric models in predicting drug response.

Main Methods:

• Analysis of genomic alterations and anticancer drug response profiles in cancer cell lines.
• Modeling of perturbed interactions within signaling pathways due to mutations.
• Development of machine learning classifiers using interaction-centric and gene-centric features.

Main Results:

• Identified novel interaction/drug response associations for afatinib and ixabepilone, not found with gene-centric methods.
• Interaction-centric classifiers significantly outperformed gene-centric classifiers in predicting responses to 164 drugs (p=2.0 × 10⁻³).
• The network approach revealed genomic alteration-drug response associations missed by gene-centric analyses.

Conclusions:

• An interaction-centric approach provides a more comprehensive understanding of cancer drug response than gene-centric methods.
• Incorporating signaling pathway interactions is crucial for advancing precision oncology.
• This network-based strategy enhances the discovery of targeted therapies and improves predictive accuracy.