An Ensemble Based in Silico Prediction of Saccharomyces Cerevisiae Proteins under Mitochondrion Organization

Protein function annotation is vital for identifying disease causative factors and for solving mysteries behind biological system complexities. As manual annotation requires costly and laborious in vitro methods, in silico protein function prediction is preferred nowadays. According to literature, one in five yeast mitochondrial proteins are known to be human disease related. This paper presents a genetic algorithmically weighted heterogeneous data ensemble to classify Saccharomyces cerevisiae proteins under 'mitochondrion organization'(GO:0007005) function defined in Gene Ontology. It consists of five euclidean-distance based nearest neighbour models and three affinity-based neighbourhood models, utilizing protein properties data, four gene expression datasets and protein interactions. The overall prediction is the weighted average of the posterior probability outputs given by the base models. Weights are determined by the standard Genetic Algorithm. The constituted base models show a fair agreement and improve the best performing base classifier by ~ 14.3%.