Topological characterization of housekeeping genes in human protein-protein interaction network

Abstract

Human housekeeping genes (HKGs) are widely expressed in various tissues, which involve in cell maintenance or sustaining cell function, and are often taken as experimental control and normalization references in gene expression experiments. Based on literature curation and up-to-date databases, we construct a large-scale human protein-protein interaction network (HPIN) and a HKGs subnetwork. Through the topological features of HKGs in the HPIN, we characterize the topological features of human HKGs. Our results indicate HKGs are with very large average degree, k-shell, betweeness, semilocal and eigenvector centralities, clustering coefficient, closeness, PageRank and motif centrality, which are all higher than that of the HPIN. Among the nine indexes, HKGs are with the average betweeness about 7 times larger than that for the HPIN, but they are also with the largest coefficient of variant (CV). The closeness of HKGs is with the smallest CV and very large median. Based on ROC analysis, we find most of the indexes and their compositions can be used to predict HKGs, with prediction accuracy around 80%. Especially, the prediction accuracy of the closeness can achieve as high as 82.36%. The investigations shed some lights on the characterization and identification of human functional genes, which have potential implications in systems biology and networked medicine.