Research On Protein-Protein Interaction Prediction And False Positive Reduction On PPI

The prediction of protein interaction is one of the most important issues in post-genomic era. Although there has been a remarkable line of research in the study of protein interaction dataset with high reliability, limited by existed technologies and predict methods, these methods to predict protein interaction do not have a good prediction performance.False positive predictions cannot be avoid in both of experimental methods and computational approaches of protein interaction prediction, therefore false positive reduction is a fundamental step to generate high reliable protein interactions. In this paper, we calculated the similarity of proteins by the semantic similarity of GO (Gene Ontology) terms which belonged to the associated proteins. Filtering algorithm is proposed based on the similarity of proteins which was used to filter the false positive predictions. The experimental results show that the proposed filtering method can decrease the false positive effectively which has enhanced the accuracy of the protein-protein interaction prediction.In this study, we use a combination of two genomic features that include protein structure information and GO annotations information, and take the model MLE to predict protein-protein interactions in Saccharomyces cerevisiae. Compared with protein interaction data sets from single information, the experimental results show that this combination method can increase the true positive fractions of the datasets drastically with only a slight increase in the false positive, which means that our method has a good prediction performance. This method can be used to assign a likelihood ratio, corresponding to the reliability to each protein interaction.