B0923
Title: A functional regression model for determining drug-response relationship in cancer genomics
Authors: Juhyun Park - ENSIIE (France) [presenting]
Evanthia Koukouli - Lancaster University (United Kingdom)
Frank Dondelinger - Lancaster University (United Kingdom)
Abstract: Cancer is a complex disease caused by abnormal cell growth in the organism. Despite its diversity in appearance, it is suspected that there may be some common biological processes that govern the underlying dynamics. Experiments using cell lines derived from patients offer valuable resources in this regard. We investigate the effectiveness of six anticancer compounds applied to different cancer cell lines under different dosage levels, adjusting for the genetic profiles of individual subjects under treatment. We assume that out of tens of thousands genes regulating proteins composition only a small proportion is actually associated with cancer cells survival in a dosage-dependent manner. Using the longitudinal and transcriptional data from the Genomics of Drug Sensitivity in Cancer (GDSC) project, we build a dose-varying regression model, a type of functional regression model with functional responses. The covariates include dose-invariant factors but their effects are assumed to vary smoothly over the dosage levels. Due to the dimensionality of the covariates, we combine a nonparametric screening with the selection of variables using a penalised regression. We evaluate the effectiveness of our method using simulation studies, focusing on the choice of the thresholds for screening and penalty parameters.
