TIGP (BIO)—Deep Learning for Predicting the Response to Chemical and Genetic Perturbations of Cancer

Abstract

The advances of genome sequencing and high-throughput screening have led to large-scale data resources for cancer discovery, such as The Cancer Genome Atlas (TCGA) and the Cancer Dependency Map (DepMap). Due to data heterogeneity and dimensionality, however, it remains challenging to comprehensively integrate these datasets to study the central dogma of pharmacogenomics: how multi-omics determine cellular response to perturbations. Our research focuses on the development of cutting-edge deep learning models to capture and predict intricate pharmacogenomic patterns among high-dimensional genomics and high-throughput chemical and genetic screens. The talk will introduce several of our pioneering models that accurately predict cancer cells’ i) response to a broad panel of approved and investigational anti-cancer drugs and ii) genetic dependencies on potential cancer genes. The models feature specialized ‘transfer learning’ designs that enable the translation of in vitro screens to impracticable-to-screen tumors. In addition, we have developed a suite of user-friendly web tools to facilitate access to large-scale pharmacogenomic data and deep learning models, thereby promoting the utilization by biomedical and clinical researchers. The studies demonstrate the exciting promise of deep learning for precision oncology by implementing an intelligent prioritization of chemical and genetic targets to enhance the efficiency and precision of drug discovery and development.
 
Lab website: https://chiu-lab.org/

► Please click Webex Meeting Link to join the talk.

Download