BME 280B Seminar: Computational Models of Biological Systems
Presenter: Chen-Hsiang Yeang, Associate Research Fellow, Institute of Statistical Science of Academia Sinica
Description: Computational models are roughly categorized into two types: describing the patterns of the phenomenon or data (description-driven models) and explaining the phenomenon or data with simpler, comprehensible rules (explanation-driven models). When building a model, the choice of the mixture ingredients of these two classes depends on the nature of the problem, availability of the knowledge and data of the underlying system. In this talk, I will give an overview of five models with varying combinations of descriptive and explanatory elements on different biological problems. First, I will introduce a joint model to capture irreversible and reversible drug resistance mechanisms of cancers and a dynamic treatment strategy to tackle drug resistance. Second, I will introduce a backward deconvolution algorithm based on probabilistic graphical models to unravel the cell type heterogeneity of the RNASeq data. Third, I will introduce a deep neural network model to integrate direct and indirect associations of genotypes and images with phenotypes. Fourth, I will introduce an experimental and computational framework to predict protein stability and discover motifs at C-terminals. Fifth, I will introduce three algorithms to demarcate independent holes of specified dimensions in large networks. These works illustrate how “model selection” should be tailored for specific biological problems.
Bio: Chen-Hsiang Yeang is currently an associate research fellow at the Institute of Statistical Science of Academia Sinica. His research interests focus on several areas in computational biology and data science: 1) cancer genomics, 2) cancer treatment, 3) molecular evolution, 4) network topology, and 5) machine learning.
Hosted by: Professor Josh Stuart, BME Department
Room: PSB-240
