Advancing Drug Development Through Strategic Cell Line and Compound Selection Using Drug Response Profiles

TL;DR

This study shows that drug-response profiles are more effective than omics data for predicting drug sensitivity in cancer cell lines, aiding early drug development and biomarker discovery.

Contribution

The paper introduces a predictive approach using drug-response profiles that outperform omics features, with validation through explainability and application to breast cancer cell lines.

Findings

DRP descriptors outperform omics data in prediction accuracy

Identified known and novel biomarkers for drug sensitivity

Successfully distinguished drug responses between MCF7 and MCF10A cells

Abstract

Early identification of sensitive cancer cell lines is essential for accelerating biomarker discovery and elucidating drug mechanism of action. Given the efficiency and low cost of small-scale drug screens relative to extensive omics profiling, we compared drug-response panel (DRP) descriptors against omics features for predictive capacity using gradient boosting tree models across the GDSC and CCLE drug response datasets. DRP descriptors consistently outperformed omics data across key performance metrics, with variable performance across different drugs. Using complementary explainability approaches, we confirmed known MAPK-inhibitor sensitivity signatures, and identified novel potential biomarker candidates for MEK1/2 and BTK/MNK inhibitors. Lastly, to demonstrate the utility of this approach in distinguishing phenotypes, we applied our models to the breast cancer line MCF7 versus the non-tumorigenic MCF10A, and successfully identified compounds that selectively inhibit MCF7 while sparing the non-tumorigenic MCF10A. This methodology, developed using focused drug and cell line panels, supports early-stage drug development by facilitating rational cell line selection and compound prioritisation, enabling more efficient biomarker identification and candidate assessment.