文摘
With the huge amount of biological data produced by recent high-throughput technologies such as ChIP-chip, ChIP-Seq and RNA-Seq, it is essential to use computational methods for data analysis. This thesis developed a series of computation tools to mine critical biological knowledge from large-scale genomic data. Those tools were used to characterize the relationships between regulatory factors and gene expressions across development stages, to uncover the regulation functions from combinatorial bindings of factors, and to discover new genomic annotations. The contributions of this thesis lie in several aspects. First, we developed a unique peak calling method that accurately distinguished adjacent peaks and obtained the correct number of summit positions. Second, we proposed two effective metrics for measuring gene activeness using RNA-Seq data. Third, we enhanced a standard data visualization technique in several ways. Fourth, we conducted a study on two important factors, which led to the categorization of known genes, and a novel approach for new gene prediction. Fifth, we studied the combinatorial bindings of multiple factors, uncovered the most frequently occurring factor sets, and calculated scores to indicate co-binding significances. We also examined the enrichments of factor sets over various types of genomic segments. Last, we mined protein-DNA interactions from positive and unlabeled data using a classification approach.