Quantifying the dependency between biomolecular composition and downstream cellular phenotypes is a fundamental open problem in biology. Advances in multimodal single cell measurement technologies provide an opportunity to apply new computational frameworks to dissect the contribution of individual and combinations of biomolecules to a given phenotype. In this work, I analyzed complex relationships related to transcriptional activity, receptor abundance, apoptotic decisions, and Ca2+ signaling. In the first chapter, I used an information theory approach to analyze multimodal data of the expression of 83 genes in the Ca2+ signaling network and the dynamic Ca2+ response in the same cell. We found that the overall expression levels of these 83 genes explain approximately 60% of Ca2+ signal entropy. The average contribution of each single gene was 17%, revealing a large degree of redundancy between genes. In the second chapter, I connect apoptotic decision-making to transcriptional activity, receptor abundance, and caspase activity. In both approaches, I quantify relationships between cell state and phenotype from multimodal, high-dimensional data.

The recent progress of single cell quantitative microscopy produced a plethora of data that demonstrated cell to cell variability in signal transduction, which provoked many possible theories that sought to explain the phenomenon of cellular variability. In order to investigate the mechanism of cell to cell variability, mathematical modeling needs to be combined with experimental approach to gain systems-level understanding. However, until now there have been few studies which applied mathematical modeling to study single cell data because of the numerous practical and theoretical hurdles. Here I develop a computational toolbox specifically designed to perform single cell model fitting efficiently on a population scale by utilizing big data platform. The toolbox fits time series data using Bayesian parameter optimization that produces a parameter distribution for the mathematical model. The model fitting is carried out on the Google Cloud Platform which scales well with the increasing number of single cell data and allows for fitting a population of single cell data at fast pace and extreme low cost. I combine the same model fitting algorithm with single cell fluorescent microscopy experiment to study cell to cell variability in calcium signaling pathway. By analyzing the parameter distributions from single cell fitting using clustering methods, I discover distinct clusters within the parameter distributions of cell population that vary the most at IP3 receptor within the calcium signaling pathway. The theoretical findings are corroborated by subsequent experiments which demonstrated that IP3 receptors played significant role in defining calcium response identities in cells.

The human body is a patchwork of tissues working collectively to maintain homeostasis and achieve fitness objectives. Within these tissues are cells, the smallest units of life, which produce a tissue’s morphology through emergent behavior defined by a specific pattern of gene expression. However, the observation and interpretation of this behavior clash, creating serious contradictions that need to be resolved. Isogenic populations of cells exhibit significant phenotypic heterogeneity but are still treated as monolithic entities exhibiting homogenous behavior under equilibrium. Holding this contradiction together are theories of cell behavior that posit stochastic fluctuations as responsible for heterogeneity, treating it as different snapshots of a memoryless, dynamical, process. To test this theory, we’ve tested the persistence of distinct cellular states, such as the allele-specific variability of an exogenous reporter system and the intracellular calcium response to ATP, finding a lack of ergodicity. For the former we have also measured the co-fluctuation of underlying chromatin states, linking expression variability to variability in histone modifications. This work challenges existing paradigms of cellular heterogeneity, implying deeper regulation of cell state and more functionally stratified cell biology.

Chromosomal position effect, also known as position effect variegation, has been extensively studied for almost a century. A systematic approach to study positional effect is to isolate genetic from epigenetic factors specifically to measure the expression of the same gene positioned in different chromatin contexts. Current strategies to target a reporter gene at multiple genomic locations are not capable of increasing both the sensitivity and throughput of data. Here, we developed a new massively parallel method to create and identify isogenic reporter clones. This method allowed us to interrogate the effect of chromatin environment on gene expression variability and epigenetic drug sensitivity as well as identify their underlying mechanisms. In human cells, we found that the protein expression mean and noise significantly are varied by the genomic location of the gene. By mapping our measurements of reporter expression at different genomic loci with epigenetic profiles of the transcription factor enrichment and the distance to chromatin states, we identified the factors that impact gene regulation. Some factors are involved in mediating both gene expression mean and noise, while other only control one of these features. Moreover, we discovered wide-spread loci-specific sensitivities to epigenetic drugs for three distinct chemical compounds that target histone deacetylase, DNA methylation and bromodomain proteins. By leveraging ENCODE data on chromatin modification, we identified features of chromatin environments that are most likely to be affected by these epigenetic drugs.

Wound healing begins as soon as the wound occurs and can continue for days, weeks, or even months. Although decades of research have established the multiple-steps required to heal a wound, little is known about the cellular response immediately following the wound. Furthermore, because the location of the wound is relevant to the healing process, healthy cells responding to a wound require positional information, or knowing where they are in relation to the wound. Thus, there are two important questions not fully understood in wound healing: 1. How do cells know where they are? and 2. How do the initial signaling mechanisms influence downstream genetic expression and overall healing? Here, using recent technological advancements in single-cell fluorescent microscopy and fluorescent biomarkers, we determine that ATP, an initial signaling molecule released from wounded cells, creates spatial patterns immediately following an epithelial wound using a simple Release and Diffusion mechanism. Then we determine that epithelial cells use paracrine signaling to spatially average the initial wound response signals over a specific distance to maximize the signal-to-noise ratio. Finally, we establish a method titled SpaSeq (Spatial Sequencing) that utilizes FACS and RNA-Seq in addition to fluorescent proteins to measure the spatio-temporal gene expression patterns following a wound.