Integrative analysis of transcriptional regulation using computational approaches.

Abstract

Transcriptional regulation is one of the most important means for cells to control its gene function. The availabilities of genome sequences and high-throughput technologies, like gene expression and ChIP-chip, have enabled scientists to study gene expression and protein-DNA binding on a genomic scale. These resources and technologies facilitate transcription regulation studies but also raise challenges in data analysis. Although great achievements have been made through various computational methods, many questions still remain in computational transcription analyzes. The thesis addresses key issues in computational transcription analyses: epigenetic regulation, data integration and biological hypothesis generation. First, a Bayesian classifier model discriminates high occupancy MYC binding sites from low occupancy sites. The model incorporates sequence attributes as well as epigenetic attributes. It shows that epigenetic features are important for MYC/DNA binding, but they are ignored in previous MYC binding prediction studies. Second, the MYC binding prediction is combined with gene expression analysis from cancer studies and gene ontology analysis to identify MYC target genes. The result shows that the integration of multiple data sources improves the prediction accuracy. Finally, I analyze a microarray dataset is analyzed to identify candidate genes and transcriptional mechanisms which mediate insulin resistance in adipocytes treated by growth hormone. Gene Set Enrichment Analysis (GSEA) identifies several pathways that are relevant to insulin resistance, including insulin signaling which contains two upregulated genes SOCS2 and PIK3R1. The Chinese Restaurant Clustering (CRC) algorithm identifies a cluster that includes SOCS2, PIK3R1 and several transcription factors like BCL6. Promoter analysis predicts conserved transcription factor binding sites in multiple genes in this cluster. A potential mechanism is proposed for growth hormone-induced insulin resistance in adipocytes involving deregulation of BCL6 and STAT5 by growth hormone leading to high expression of SOCS2 and PIK3R1, which may contribute to insulin resistance in this study. This thesis explores the integration of various methods and high-throughput datasets to decipher transcriptional regulation in different studies. The work shows that using integrative strategies and applying appropriate computational approaches are effective and fruitful in the study of transcriptional regulation.