My current research focuses on stochastic modeling of gene expression, in particular rare events, and on developing novel analytical and computational approaches for entropy-regularized reinforcement learning.
One of the fundamental problems in biology is the identification of molecular mechanisms that give rise to phenotypic variation in a population of genetically identical cells. Recent research indicates that such non-genetic individuality can arise due to intrinsic stochasticity in the process of gene expression. Correspondingly, there is a need to develop a framework for quantitative modeling of stochastic gene expression and its role in cellular processes. Research in my group has focused on combining analytical approaches from mathematics and physics with numerical simulations to model stochasticity in gene expression and its regulation.
Rare events leading to phenotypic variation in clonal cells is a recurring theme in current biological research. Prominent examples of such processes include latency in HIV-1 viral infections and reversible drug tolerance in sub-populations of cancer cells. Recent developments in nonequilibrium statistical mechanics, based on the theory of large deviations, have led to a general framework for rare-event statistics in diverse systems. My group has worked on integrating these recent developments with approaches from queueing theory to derive general results for rare event statistics in stochastic models of gene expression.
The interface between machine learning and statistical physics has given rise to some of the most significant developments in current research in artificial intelligence. A less explored area is the connection between problems in machine learning and concepts and approaches from non-equilibrium statistical mechanics. My group is currently developing new approaches for entropy-regularized reinforcement learning using the analytical framework of large deviation theory as developed in recent research in non-equilibrium statistical mechanics.