CS&E Profile: Chad Myers

Computational biology, functional genomics, machine learning

Ph.D. 2007, Computer Science, Princeton University

B.S. 2002, Computer Engineering & Math, Southern Methodist University

B.A. 2002, Physics, Southern Methodist University

About

Assistant Professor Myers joined the department in January of 2008 and specializes in computational biology and functional genomics. He has co-authored fifteen refereed publications for various conferences and journals.

Myers’s research focuses on machine learning methods for integrating large-scale genomic datasets to understand gene function and biological networks.

Research

Recent developments in biotechnology have enabled quantitative measurement of diverse cellular phenomena. For instance, microarray technology allows biologists to measure the expression of all genes in the genome on a single chip. Other technology allows high-throughput measurement of physical interactions between proteins, which are an important mechanism behind most cellular processes. These recent developments have generated an unprecedented amount of data for several different organisms. These data promise to revolutionize our understanding of biology, but integrating information across several noisy, heterogeneous datasets to derive holistic models of the cell requires sophisticated computational approaches.

My research focuses on machine learning approaches for integrating diverse genomic data to make inferences about biological networks. The main purpose of our work is to further our understanding of gene function and how genes or proteins interact to carry out cellular function. A few of my current focuses are described below.

Data integration and network inference

The main focus of my lab is methods for predicting biological networks from diverse genomic data. The goal of these approaches is to integrate information from huge repositories of genomic data including gene expression, sequence, protein localization, and protein-protein and genetic interaction data. My previous work in this area includes the development of a general system for integration of yeast genomic data, which consists of a Bayesian framework and a web-interface that allows expert biologists to intelligently browse through these collections of otherwise noisy data. This system, bioPIXIE, has been used extensively by biologists in the yeast community and has aided the characterization of several new genes.

Iterative computational-experimental methods

Directed application of high-throughput experimental technology can dramatically increase the efficiency at which we discover new biology. We are developing methods to enable this targeted investigation, and in particular, support iterative computational-experimental refinement of network models. Such methods must both predict accurate models from genomic experiments and also suggest the next experiment that removes the most uncertainty from the model.

I am currently involved in two projects in yeast demonstrating the utility of this type of approach: one which in we are using data integration to efficiently map the yeast genetic interaction network, and the other a large-scale investigation of genes involved in mitochondrial function. Both of these projects involve several months of experimental investigation that were invested based entirely on computational direction.

Analysis of genetic interactions in Saccharomyces cerevisiae

Another focus of my lab is understanding genetic interactions in the Baker’s yeast, Saccharomyces cerevisiae, and using them to make inferences about gene function. Genetic interactions are broadly defined as combinations of mutations that result in surprising phenotypes. In yeast, recent technology has enabled very fast combinatorial perturbations, producing genetic interaction measurements for millions of pairs of genes. We generally know that a gene’s interaction profile is highly informative of its role in the cell, but translating these millions of high-throughput measurements into accurate network models is challenging. We are attacking this problem with a variety of computational tools.