GENE FUNCTION PREDICTION BY THE MULTI-LAYERED CLASSIFIER WITH MULTI-FEATURES

Gene Ontology (GO) is a controlled vocabulary to describe the gene function. Each GO term is constructed by ahierarchical structure of gene function, so it is suitable for describing relationships of gene functions. We applied Bayesiannetwork model as a training model using GO terms to identify unknown gene functions, and used Bayesian network modelwith three different heterogeneous data sets and multi-layered classifier to automatically predict gene functions. This proposedmodel is comprised of a base-classifier and a meta-classifier. The base-classifier serves a base of meta-classifier withBayesian network model and meta-classifier plays role of classifying the designated GO term from the root node. Acomparative analysis of our suggested model and other gene functional annotation systems shows that our modeloutperforms than others especially in terms of a number of correctly predicted proteins.