Method for Essential Protein Prediction Based on the Naïve Bayesian Classifier and Bioinformation Fusion

Abstract

In this article, a novel approach called EPNBC has been proposed by combining the biological information with the naïve Bayesian classifier and PageRank algorithm to predict potential essential proteins. In EPNBC, the naïve Bayesian classifier is used to process the original PPI network, and a new protein interaction network with more interaction relationships is obtained. Then, two similarity matrices were obtained by using Gaussian interaction profile kernel similarity based on the protein interaction relationship and gene expression data, and a weighted protein interaction network was obtained. The improved PageRank algorithm was used to score the nodes in the network and output the protein scores in descending order. Experimental results showed that EPNBC was superior to dozens of other methods in identifying essential proteins.