Predicting Cancer Drug Response Using a Recommender System

Ms Chayaporn Suphavilai
Dr Wong Lim Soon, Kithct Chair Professor, School of Computing
Dr Nagarajan, Niranjan, Adjunct Associate Professor, School of Computing

  30 Apr 2019 Tuesday, 03:00 PM to 04:30 PM

 Executive Classroom, COM2-04-02


As we move towards an era of precision medicine, the ability to predict patient-specific drug responses in cancer based on molecular information such as gene expression data represents both an opportunity and a challenge. In particular, methods are needed that can accommodate the high-dimensionality of data to learn interpretable models with the goal of providing the right drug for the right patient at the right time.

We propose a method based on ideas from recommender systems (CaDRReS) that predicts cancer drug responses for unseen cell lines/patients based on learning projections for drugs and cell-lines into a latent pharmacogenomic space. Comparisons with other proposed approaches for this problem based on large public datasets (CCLE, GDSC) show that CaDRReS provides consistently good models and robust predictions even across unseen patient-derived cell line datasets. Also, analysis of the pharmacogenomic spaces inferred by CaDRReS can be used to understand drug mechanisms, identify cellular subtypes, and characterize drug-pathway associations.

Finally, we approach intra-patient drug response heterogeneity by proposing a modified version of CaDRReS for single-cell data to investigate intra-patient drug response heterogeneity, using head and neck cancer as a case study. We showed that systematically combining cell-type specific drug response predictions provided better concordance with in vitro drug response when comparing to prediction based on bulk gene expression. Furthermore, to transfer our in silico prediction to a clinic, we incorporate clinical drug response information to predict an upfront patient-specific drug combination that could inhibit multiple cell types identified within a patient, resolving intra-patient heterogeneity.