Bioinformatics Seminar
Time: 11AM
Venue: Davis Auditorium and Online
10 November 2026
Evaluating signature matching with LINCS data for drug repurposing
Hengxin PanWEHI
Drug repurposing matches disease gene expression signatures against large perturbation libraries to identify compounds that reverse the disease state. The Library of Integrated Network-based Cellular Signatures (LINCS) provides an extensive collection of drug and genetic perturbation profiles for these comparisons. Standard signature-matching metrics generally treat genes as independent, often leading to uncalibrated significance estimates. Drawing on the principles of competitive gene set testing from CAMERA, my work introduces inter-gene correlation adjustments into the signature-matching framework. Using genetic perturbations with known targets as a benchmark, I show that accounting for correlation markedly improves empirical calibration and retrieval stability compared to standard approaches. I will share what these evaluations tell us about common scoring methods, and discuss practical considerations when applying these ideas to large-scale compound libraries.