Abstract

Our bodies have a remarkable capacity to repair and self-renew through tissue stem cells. How tissue stem cells make cellular decisions by coordinating the abundance and activity of thousands of RNAs and proteins to regulate tissue function is not well understood.

Gene expression can be controlled by how the DNA is packaged inside the cell (called chromatin). For example, compacting DNA can turn specific genes off while decompacting and turn specific genes on. During stem cell differentiation, the chromatin changes to support the new expression state. We recently developed a high-throughput method to profile this DNA packaging in single cells.

In this talk, I will focus on the goals and challenges of analyzing this data. In particular, I will present machine learning strategies that model the sparse data and project onto lower-dimensional manifolds in order to characterize the chromatin state along stem cell differentiation. I will also discuss systems biology methods that enable physically meaningful interpretations of the data in terms of protein regulators that underlie changes in chromatin state.

Overall, combining machine learning methods with systems biology creates maps of cellular differentiation and provides interpretable tools to understand and probe these maps.