PROJECT SUMMARY / ABSTRACT The goal of my research program is to understand how transcription factors (TFs) direct the regulatory programs that underlie cell fate decisions. My lab currently focuses on a fundamental step in TF regulatory activity: how do newly induced TFs establish their DNA binding patterns? TFs should have binding affinity for millions of sites along the typical vertebrate genome, yet only a small fraction appears to be bound in a given cell type. Moreover, the cohort that are bound changes across cell types and developmental timepoints. We have developed pioneering machine learning approaches for characterizing regulatory genomic events and understanding TF binding specificity. We have collaboratively applied our computational approaches to understand cell fate decisions in cell differentiation systems, finding new ways in which the binding of induced TFs can be influenced by preexisting chromatin environments. This proposal aims to integrate algorithmic development and applied analysis of regulatory systems to gain a comprehensive understanding of how genome-wide TF binding patterns are predetermined by chromatin regulatory states. While many have cataloged the concurrent chromatin features that coexist with TF binding sites in a static context, this proposal focuses on the dynamic settings that are typical of cell fate decisions. How does the chromatin landscape in a given cell type shape where a newly induced TF will bind? Theme 1 will continue our development of machine learning methods for studying dynamic TF binding activities. We will focus on novel neural network architectures that can separate sequence and chromatin features to explain induced TF binding patterns. Drawing on our unique expertise and methodologies, we will ask whether integrating 3D genome organization or protein-DNA binding subtype modes (e.g., direct vs. indirect DNA binding) can explain why certain sites become bound by induced TFs. We will further ask if DNA binding predeterminants are transferrable: can we predict where a given TF will bind if introduced into a new cell type? Theme 2 will analyze how TFs interact with established chromatin environments during cell fate decisions. We will ask how paralogous Forkhead box TFs recognize distinct binding targets, even when they have similar DNA binding preferences and are expressed in the same chromatin environment. To understand how TF binding sites and regulatory activities can change as cells proceed down differentiation trajectories, we will continue long-standing collaborations that examine chromatin-dependent TF regulatory behaviors during neuronal subtype specification and hematopoiesis. Complementary to these efforts, we will build integrative regulatory models of temporal chromatin accessibility dynamics at the single cell level. The two themes will synergize to provide the computational tools and applied analyses that will enable a more complete understanding of TF regulatory specificity during cell fate decisions.

PROJECT NARRATIVE Transcription factor proteins bind to the human genome and regulate the genes that govern human health. Understanding how transcription factors find their regulatory targets on the genome informs us of how the diversity of cell types in the human body arises during development and how cells go wrong during disease. This project aims to develop machine learning tools for understanding how transcription factors find their DNA binding sites when they are expressed in cells.