PhD Proposal: Adaptive Evolution of Continuous Traits and Interaction Modeling in Species and Tumors: Methods and Biology
MG Hirsch
Wednesday, December 4, 2024, 1:00-3:00 pm 
Abstract
Abstract:
Evolution is the change in genomic and phenotypic traits over long periods of time. These changes can be random, resulting in neutral evolution (genetic drift), or they can be caused by selective pressures in the environment, resulting in constrained evolution (purifying/negative selection) or adaptive evolution (disruptive/diversifying selection). Selective pressures cause certain phenotypes to have higher fitness, meaning that individuals with those traits are more likely to reproduce. This causes the more fit phenotypes to spread through the population. While selective pressures act on the phenotypic level, phenotypes rely on a variety of genetic and epigenetic mechanisms. When a certain phenotype is selected for, so are the underlying mechanisms causing it. Here, I present a variety of methods to study different facets of adaptive evolution of continuous traits, specifically gene expression, with application to cancer tumor and species evolution.
First, I present a study that uses stochastic modeling of gene expression evolution to evaluate the adaptation of gene expression of tumor subclones with varying phenotypes. Gene expression heterogeneity can cause tumors to be more difficult to treat with certain therapies, and studying the selective forces behind this heterogeneity can enable the design of more effective treatment methods. By applying Ornstein-Uhlenbeck methods to gene expression of different subclones, I show that subclones with different phenotypes have genes with adaptive expression associated with contrasting cellular processes.
Second, I present a project implementing a method for the mathematical modeling of interactions between tumor subclones that have adapted to have different phenotypes. Studying how subclones interact with each other and the immune system can provide further insights into their adaptation and assist in creation of targeted treatment. I use a game-theoretic approach based on Lotka-Volterra equations to analyze the tumor growth and the type and magnitude of subclone-subclone and subclone-immune system interactions.
Finally, I propose a project implementing a method for the estimation of adaptive co-evolution of continuous traits, like gene expression. Although it is known that gene expression does not evolve independently, most methods predicting adaptation assume independence. This is in part because of computational difficulties when trying to allow for shared parameters between traits. In this project, I intend to use multivariate Ornstein-Uhlenbeck processes to predict co-adaptation of pairs of traits and use those results to build a network to represent co-adaptation of larger sets of traits.
First, I present a study that uses stochastic modeling of gene expression evolution to evaluate the adaptation of gene expression of tumor subclones with varying phenotypes. Gene expression heterogeneity can cause tumors to be more difficult to treat with certain therapies, and studying the selective forces behind this heterogeneity can enable the design of more effective treatment methods. By applying Ornstein-Uhlenbeck methods to gene expression of different subclones, I show that subclones with different phenotypes have genes with adaptive expression associated with contrasting cellular processes.
Second, I present a project implementing a method for the mathematical modeling of interactions between tumor subclones that have adapted to have different phenotypes. Studying how subclones interact with each other and the immune system can provide further insights into their adaptation and assist in creation of targeted treatment. I use a game-theoretic approach based on Lotka-Volterra equations to analyze the tumor growth and the type and magnitude of subclone-subclone and subclone-immune system interactions.
Finally, I propose a project implementing a method for the estimation of adaptive co-evolution of continuous traits, like gene expression. Although it is known that gene expression does not evolve independently, most methods predicting adaptation assume independence. This is in part because of computational difficulties when trying to allow for shared parameters between traits. In this project, I intend to use multivariate Ornstein-Uhlenbeck processes to predict co-adaptation of pairs of traits and use those results to build a network to represent co-adaptation of larger sets of traits.
Bio
MG Hirsch is a 4th year PhD student advised by Teresa Przytycka at the National Institutes of Health and Erin Molloy at the University of Maryland studying computational biology with applications to cancer.
This talk is organized by Migo Gui