Computational Analysis and Modeling of High-Throughput Data to Understand T-helper Cell Differentiation

T-helper cells are essential for adaptive immunity. During immune response T-helper cells are influenced by cytokines which steer cell differentiation into cellular subsets having specific functions. The process is affected by cellular subsystems and causes profound changes in epigenetic modifications as well as gene and protein expressions which can be experimentally observed using high-throughput technologies. This thesis has three objectives. 1) To identify and characterize molecular elements involved in T-helper cell differentiation and immune response through analyzing datasets using bioinformatic tools. 2) To develop computational tool to detect enrichments of trait associated single nucleotide polymorphisms (SNPs) on genomic regions. 3) To develop computational frameworks for characterizing dynamic models for regulatory networks. The goals have been achieved in five studies. In the first study, proteomes and transcriptomes of Th17 and iTreg cells were profiled and analyzed to understand how they change during early phases of cell differentiation and how the transcriptomes and proteomes of the cell types differ from each other. The second publication characterized bindings of transcription factor (TF) STAT3 genome-wide during Th17 cell differentiation. As a SNP can alter binding affinity of a TF, we investigated whether SNPs associated with immune diseases co-localize in STAT3-binding sites. The analysis applied publicly available information on SNPs and empirical statistical methods. The third publication proposes a computational tool snpEnrichR implemented in R language for facilitating co-localization analyses of SNPs and genomic regions. Co-localization analysis of SNPs associated to various traits and STAT6 binding sites of cells differentiating toward Th2 type showed that incorporating proxies of the tag-SNPs enhances co-localization detection. The fourth publication introduced a method to infer dynamically evolving regulatory networks from time-course data. The method couples mechanistic ordinary differential equation (ODE) models with a latent process that approximates the network structure rewiring process. When applied to Th17 RNA-seq data the method predicted lineage specific subnetworks that are activated sequentially and control the differentiation process in an overlapping manner. The fifth publication studies the dynamic interplay of histone modifications signaling enhancer activity and transcription factor binding modeling them using systems of ODEs and simulated time-course data focusing on the parameters of the models and model selection. The method is able to find the correct model when measurement noise level is reasonable and the number of measurement time points is adequate. The datasets generated and the analyses performed as part of this thesis help to understand of T-helper cell differentiation better. The developed computational frameworks and tools available for use as well as further developments. The results provide a valuable resource for the community.