Identifying The Modular Structures In Protein Interaction Networks

Identifying module structures of biological systems is an important issue to understand the biological system's organization and function. In order to identify modules with high biological relevance, the degree concept was extended from single vertices to subgraphs; and a formal and flexible definition of a module in a network was proposed. Combing the new module definition and a new agglomerative algorithm, a computational tool, ModuleSpider, was developed to recognize modules within a network. Applying ModuleSpider to the yeast core protein interaction network from the DIP database identified 97 simple modules which are significantly biological meaningful according to similar GO terms of proteins in the modules. A comparison between the ModuleSpider modules and modules defined by Radicchi et al. indicates that proteins in ModuleSpider modules show stronger GO term co-occurrences. Further, the ModuleSpider is able to construct the interaction web of modules. The graphical presentation of the interconnection web between modules provides insight to the high level relationships of different functional modules.In addition, the sequence similarity searching algorithms are studied. Based on the thorough analysis on Blast, a Blast-like fast sequence alignment tool– MyBlast, was proposed and implemented. And, the core MyBlast algorithm has been successfully applied in the True EST alignments and exon regions of gene sequences identification program– EDSAc. When tested with human gene sequences in the standard dataset HMR195 and evaluated with the standard measures of gene prediction performance, EDSAc outperform that of the counterpart TAP in specificity sensitivity at the nucleotide level.