Predicted Molecular Interaction Network And Network Based Systems Biology Analyses

Molecular interactions are essential for almost all cellular processes. Deciphering the molecular interaction network not only provides insights into protein functions but also advances our understanding of higher-level phenotypes and their regulation. In Saccharomyces cerevisiae, Homo sapiens, Drosophila melanogaster, Caenorhabditis elegans and Arabidopsis thaliana, genome-wide yeast two-hybrid screens and large-scale affinity purification/mass spectrometry studies have been reported. A number of databases, such as IntAct, BioGRID, BIND, and TAIR, have been established as repositories for interaction data. However, no experiment aiming to chart an entire plant and human interactome has been attempted. In Arabidopsis, and human, reported experimental interactions represent only a small fraction of their total interactomes.With careful selecting training and testing dataset, integrating side evidence, adjusting prediction model, we inferred human and Arabidopsis interactome. The Predicted Arabidopsis Interactome Resource (PAIR) presents149,900potential molecular interactions, which contains4,545experimental protein interactions and145,494inferred interactions. PAIR was expected to cover～24%of the entire interactome with～40%precision. And the Human Interactome Resource (HIR) included155,974interactions, which integrated69,586experimental protein interactions and90,208inferred interactions. HIR was expected to cover22.1%of the total human interactome with a per-interaction reliability of41.0%.These inferred interactions can nicely power several network topology-based systems biology analyses, such as gene set linkage analysis, protein function prediction, and identification of regulatory genes demonstrating insignificant expression changes. Especially in the case study of expression data analysis, gene set linkage analysis exhibited much superior capability, as compared to the widely used term enrichment analysis, to interpret the molecular phenotype of statin-induced selective apoptosis and suggest clues for follow-up investigations.This study demonstrates that, although PAIR and HIR still have limited coverage, it is rich enough to capture many significant functional linkages within and between higher-order biological systems, such as pathways and biological processes, and improving the capability of these analyses to integrate existing knowledge and suggest novel insights into the function and coordination of genes and gene networks.