Simultaneous Inference of Cell Types, Lineage Trees, and Regulatory Genes From Gene Expression Data

Abstract

Important goals of developmental biology include identifying cell types, understanding the sequence of lineage choices made by multipotent cells and unconvering the molecular networks controlling these decisions. Achieving these goals through computational analysis of gene expression data has been difficult. In this dissertation supervised by Sharad Ramanathan, I develop a probabilistic framework to identify cell types, infer lineage relationships and discover core gene networks controlling lineage decisions. Working with Sandeep Choubey and Sumin Jang, we infer the gene expression dynamics of early differentiation of mouse embryonic stem cells, revealing discrete state transitions across nine cell states. Using a probabilistic model of the gene regulatory networks, we predict that these states are further defined by distinct responses to perturbations and experimentally verify three such examples of state-dependent behavior. Working with Vilas Menon and Sam Melton, we infer a lineage tree for early neural development and putative regulatory transcription factors from single-cell transcriptomic profiles. The lineage tree shows a prominent bifurcation between cortical and mid/hindbrain cell types, and the inferred lineage relationships were confirmed by clonal analysis experiments. In summary, this study provides a framework to infer predictive models of the gene regulatory networks that drive cell fate decisions.