Disease Cluster Detection and Functional Characterization

Abstract

Mechanisms underlying human diseases have been revealed with the development of molecular biology. The underlying molecular basis of disorders is valuable in prevention, diagnosis, and treatment. Decade-long efforts have been devoted to investigating disease–gene association through positional cloning of disease genes and genome-wide association studies. In particular, correlations among different diseases have been discovered by many clinical cases. The shared disease-associated genes may help reveal the intrinsic relationship in the genetic level, provide an access to evaluate disease similarity, and establish a human disease network. Although many methods have been proposed to measure disease similarity, they only consider the genes or functions directly annotated to diseases but ignore the interactions among genes or functions. These interactions cause deficiency in disease classification. Basing on network-based disease module, we presented a systematic research to further investigate the relationship among different human diseases and explore whether this correlation depends on the functions of corresponding disease genes. On the one hand, a disease clustering based on the separation score between diseases is applied to divide 299 diseases into 15 relatively separated disease clusters. On the other hand, an optimal clustering scheme discriminating 15 disease clusters was learned based on disease-associated genes, their GO terms, and KEGG pathways annotations. The detected key signatures showed the highest relevance to distinguishing distinct disease clusters and represented the essential functions in corresponding pathogenesis. This study provides a novel approach to predict the network and function characteristics and reveals the functional essence of diseases.