Detection and Characterization of Patient Subgroups with Heterogeneous Behavior

The rapid developments in collecting, storing, transmitting, and managing massive amounts of data have led to unique opportunities and challenges in Statistics and the emerging field of Data Science. Motivated by large-scale medical studies where huge volumes of data with complex structure are produced, Professor Narisetty aims to develop a novel statistical framework to detect patient subgroups which exhibit heterogeneous behavior and to characterize these subgroups based on their biological and clinical features. The statistical framework developed will introduce novel statistical models, methodology to estimate the models, and computational algorithms scalable to Big Data settings. Quantile regression models and Bayesian techniques will be utilized for developing this framework.

To make the proposed statistical and computational research suitable for a variety of scientific applications, high dimensional and complex structures of the features will be accommodated. In the medical context, the research methods will be useful for personalized medicine so that patients belonging to more responsive subgroups can be treated and those belonging to subgroups with side effects can be avoided. Software packages in R will be developed to make the proposed methods for subgroup analysis accessible for wide use by scientists and practitioners.