Drug-Response Prediction from Transcriptomic Profiles

Can the RNA-seq profile of a cancer cell line predict how strongly it responds to a drug - and does the prediction hold in another laboratory's screen? This project trains per-drug elastic nets and a multi-task neural network on the GDSC2 screen (312 cell lines, 20 drugs, PharmacoDB AAC) and tests them on the CCLE, CTRPv2 and gCSI screens with RNA-seq from another sequencing centre. Expression clearly beats the tissue of origin in cross-validation by cell line (Spearman 0.48 vs 0.32), the models recover known markers such as SLFN11, EGFR, CDKN2A and MDM2, new cell lines reach 0.40-0.46 in the other screens, and for known lines the agreement with other labs is as high as GDSC2's own measurements allow.

Category: Bioinformatics - Transcriptomics and Gene-Expression Analysis

Transcriptomics & Gene-Expression Analysis Project B19 from ProTechSC Academy, with graphical abstract, video demonstration and project guidance.