Int J Biol Sci 2011; 7(1):61-73. doi:10.7150/ijbs.7.61

Research Paper

Prediction of Human Disease-Related Gene Clusters by Clustering Analysis

Peng Gang Sun1✉, Lin Gao1✉, Shan Han2

1. School of Computer Science and Technology, Xidian University, Xi'an, 710071, China
2. Faculty of Science, University of Copenhagen, Copenhagen, 1307K, Denmark

This is an open access article distributed under the terms of the Creative Commons Attribution (CC BY-NC) License. See for full terms and conditions.
Sun PG, Gao L, Han S. Prediction of Human Disease-Related Gene Clusters by Clustering Analysis. Int J Biol Sci 2011; 7(1):61-73. doi:10.7150/ijbs.7.61. Available from

File import instruction


Since genes associated with similar diseases/disorders show an increased tendency for their protein products to interact with each other through protein-protein interactions (PPI), clustering analysis obviously as an efficient technique can be easily used to predict human disease-related gene clusters/subnetworks. Firstly, we used clustering algorithms, Markov cluster algorithm (MCL), Molecular complex detection (MCODE) and Clique percolation method (CPM) to decompose human PPI network into dense clusters as the candidates of disease-related clusters, and then a log likelihood model that integrates multiple biological evidences was proposed to score these dense clusters. Finally, we identified disease-related clusters using these dense clusters if they had higher scores. The efficiency was evaluated by a leave-one-out cross validation procedure. Our method achieved a success rate with 98.59% and recovered the hidden disease-related clusters in 34.04% cases when removed one known disease gene and all its gene-disease associations. We found that the clusters decomposed by CPM outperformed MCL and MCODE as the candidates of disease-related clusters with well-supported biological significance in biological process, molecular function and cellular component of Gene Ontology (GO) and expression of human tissues. We also found that most of the disease-related clusters consisted of tissue-specific genes that were highly expressed only in one or several tissues, and a few of those were composed of housekeeping genes (maintenance genes) that were ubiquitously expressed in most of all the tissues.

Keywords: Disease-related gene cluster, Clustering analysis, PPI network, Gene expression data