Prediction of protein function using trust based cumulative strategies

Physical interactions between the proteins in a living organism helps in identification of most protein-protein interaction data. The annotated proteins are previously known by their functions. Their knowledge is definite. The un-annotated proteins are annotated based on estimation of such similar functions. Generally a cluster containing annotated nodes with their adjacent unlabeled nodes is assumed to have homogeneity of functions within. Though the interaction data are generally very noisy, a Bayesian model is presented to predict protein functions after a series of known experiments or several hypotheses over neighborhood properties are conducted or assumed. The experimental results in this effort have shown that there is a better performance in evaluation of weighted accuracy of functions over prediction of data set.