Development And Application Of A Novel Domain-Based Gene Function Prediction Algorithm

Gene function prediction remains an important issue in the post-genomics era. Accurate information of gene function is essential for the interpretation of biological mechanisms. However, current experimental methods cannot meet the needs of determining gene function in a large-scale way. Computational methods for gene function prediction thus are particularly essentialwith an increasing requirement of efficiency and accuracy. In this study, we developeda novel domain-basedgene function prediction algorithm, and assessed the predictionperformance by known GO annotations. The results indicated that our method achieved a relatively high accuracy and outperformed other domain-based methods significantly. We also proposed a novel algorithm which combined together our method and a typical sequence-based gene function prediction approach, GOtcha.The combined results further improved predicting accuracy. The combined method was also used to predict the GO annotations of the whole-genome of Mycobacterium tuberculosis. The prediction results were validated in different ways. The predicted GO annotations were further used to analyze a set of gene expression data. The functional analysis result of differentially expressed genes using predicted GO annotations indicated new possible mechanism of action of capreomycin, showing potential application value of function prediction for further biological analyses.