Unlocking Cellular Computation and Information Processing through Multidimensional Single-Cell Data

Cells are computational entities that process external signals through networks of interacting proteins and reconfigure their state via biochemical modifications of proteins and changes in gene expression. Despite progress in the understanding of signaling biology, graph diagrams typically used as depictions of signaling relationships only offer qualitative abstractions. New single-cell measurement technologies provide quantitatively precise measurements of dozens of cellular components representing important biochemical functions. However, a major challenge in deciphering single-cell signaling data is developing computational methods that can handle the complexity, noise and bias in the measurements. I will describe algorithms that quantify the flow of information through signaling interactions and mathematically characterize relationships between signaling molecules, using statistical techniques to detect dependencies while mitigating the effect of noise. I will show how these algorithms can be utilized to characterize signaling relationships in immune cells, detect subtle differences between cell types, and predict differential responses to perturbation. Next, I will analyze T cells from non-obese diabetic (NOD) mice and show that previously recognized defects in extracellular-signal-regulated kinase (ERK) signaling can be traced back to a small receptor-proximal defect that is amplified through reconvergence in the network. Then, I will show how multidimensional extensions of these techniques can be used to track dynamic changes in the relatively unknown network driving the epithelial-to-mesenchymal (EMT) transition that occurs during cancer metastasis, with the goal of predicting drugs to halt the process. Finally, I will discuss future directions involving integration of gene expression and other data types in order to gain a more complete picture of cellular computation.
I obtained my undergraduate degrees at Kalamazoo College and the University of Michigan dual major in Computer Engineering and Math (part of a 3-2 engineering program). Then, I stayed on at the University of Michigan EECS department to obtain my Ph.D. in 2008 with professors Igor Markov and John Hayes. For my PhD, I worked in VLSICAD focusing on network models to represent probabilistic effects in nano-scale logic circuits, and algorithms to analyze the reliability of these circuits. My Ph.D. work won a best paper award at DATE and an EDAA outstanding dissertation award. My thesis was published as a book by Springer. After my Ph.D., I went to IBM TJ Watson Research Center for 2 years, where I worked on formal methods to isolate errors in logic circuits and derive minimal corrections to errors in the late stages of design. My work at IBM resulted in several patents and was chosen as a division accomplishment of the year. Next, I decided to move from the realm of VLSI circuits to biological circuits and joined the computational biology lab of Dana Pe'er at Columbia University as a postdoctoral researcher. While there, I've worked on analyzing and modeling signal transduction pathways using mass cytometry data. My work has been published in Science and PNAS and other journals.