Presentation: 2017 Norman Breslow Endowed Lecture: Complex Model Building for Precision Medicine Using Summary-Level Information from Big and Disparate Data Sources
Speaker: Nilanjan Chatterjee, Ph.D., Bloomberg Distinguished Professor, Department of Biostatistics, Johns Hopkins Bloomberg School of Public Health
Abstract: Extraction of information from rich and diverse datasets through summary-level statistics, as opposed to individual level data, can be appealing because of various practical and ethical considerations. In this talk, I will describe statistical methods for building complex models using summary-level information in two distinct applications. One involves assessment of genetic architecture of complex traits by modeling of effect-size distributions using association statistics available from large genome-wide association studies. The other involves development of a generalized meta-analysis framework for unified model building using information on parameters from disparate, but possibly overlapping, sub-models fitted to different studies. Implications for future precision medicine efforts towards disease prevention will be discussed for both applications.
About Dr. Chatterjee: Nilanjan Chatterjee leads a broad research program in quantitative research that cuts across multiple areas of modern population-based biomedical science including statistical genetics/genomics, precision medicine and big data. The scientific goals of his studies include discovery of new biomarkers, understanding disease mechanisms, characterizing disease risk and developing risk-stratified approaches to disease prevention. He has extensively collaborated in recent genome-wide association studies (GWAS) that have led to identification new cancer susceptibility SNPs, provided characterization of heritability, genetic architecture and gene-environment interaction, and led to better understanding of potential for genetic risk stratification for cancer prevention. Prior to joining Hopkins, Dr. Chatterjee worked at the National Cancer Institute for 16 years and led the Biostatistics Branch of the Division of Cancer Epidemiology and Genetics during 2008-2015.
Dr. Chatterjee identifies important theoretical and methodological problems in statistics from his applied research. Motivated from genome-wide association studies, for example, he has defined novel and robust methods for analyzing genetic associations, and gene-gene/gene-environment interactions. He has provided mathematical characterization of power of polygenic risk prediction models based on underlying genetic architecture of diseases. Outside the realm of statistical genetics, he has long-term interest in developing statistical methodologies for building predictive models combining data from disparate sources. Most recently, he has shown how models can be built based on individual level data from a given study utilizing summary-level information from external big-data sources.
Dr. Chatterjee is an alumni of the UW Statistics graduate program and was mentored by Dr. Norman Breslow.