Jing Zhao; Ting-Hong Yang; Yongxu Huang; Petter Holme

“ In this study, we propose an approach to prioritize candidate genes by integrating data of gene expression level, protein-protein interaction strength and known disease genes. Our method is based only on two, simple, biologically motivated assumptions—that a gene is a good disease-gene candidate if it is differentially expressed in cases and controls, or that it is close to other disease-gene candidates in its protein interaction network. We tested our method on 40 diseases in 58 gene expression datasets of the NCBI Gene Expression Omnibus database. ”

“ We searched the Morbid map of the Online Mendelian Inheritance in Man (OMIM) database [33] and identified 359 genes associated with the 40 distinct diseases in our microarray data, in which one disease was associated with at least one gene. As listed in Table S2, the disease with the most known causing genes is cardiomyopathy, with 32 disease genes known. A total of 318 of the 359 genes were found to present in the protein-protein interaction network constructed from STRING, and these genes were used to validate our algorithm (see Table S2) . ”

“ This means that both the protein interaction network and the differential expression contain information that can be exploited in disease-gene ranking, as hypothesized. On the other hand, we see that putting Ø = 0 worsen the performance much more than putting η = 0, which suggest that there is more information for the benefit of predicting unknown disease genes in the interaction compared with the microarray data, at least with our setup. Furthermore, we were able to increase our method's performance by including partial information about known disease genes. ”

“ This category of methods is based on the assumption that the protein products of disease genes tend to be in close, in the protein interaction network, to differentially expressed genes. Karni et al. noticed that this problem as one equivalent to the set cover problem in graph theory, which is NP-complete [8] . Thus it is no wonder that large-scale protein networks can only be analyzed with approximate, greedy algorithms. ”

“ Specifically, if a group of genes shows a consistent pattern of different expression levels in sick subjects and a control group, then that gene is likely a strong candidate of playing a pathogenic role. ”

“ Our protein interaction network was constructed from the STRING database, which includes both physical and functional interactions integrated from numerous sources, including experimental repositories, computational prediction methods and public text collections. Eliminating self-interactions, this network consists of 1,032,872 interactions between 14,532 proteins of human genome, with their normalized interaction weights in the STRING database. ”

“ Abstract Many diseases have complex genetic causes, where a set of alleles can affect the propensity of getting the disease. The identification of such disease genes is important to understand the mechanistic and evolutionary aspects of pathogenesis, improve diagnosis and treatment of the disease, and aid in drug discovery. ”

“ This result suggests that the known disease genes were averagely ranked top 24.6% of the candidates by integrating gene expression levels with protein interactions, whereas they ranked top 41.8% and 25%, on average, respectively, if only gene expression data or protein interaction network were utilized. ”

“ When comparing ROC-curves of different classifiers, good curves lie closer to the top left corner and the worst case is a diagonal line that represents a strategy of random guessing. The total area under the ROC-curve (AUROC) is a measure of the performance of the classifier. The area lies in the interval [0.5,1] and larger area indicates better performance. ”

“ Jin, C-H & Zhou, T (2009) , Similarity index based on local paths for link prediction of complex networks., vol. 80, p. 046122 ↑ {{Citation KEGG pathway of Alzheimer's disease: http://www.genome.jp/kegg/pathway/hsa/hsa05010.html }} ↑ Raelson, JV; Little, RD; Ruether, A; Fournier, H & Paquin, B (2007) , Genome-wide association study for Crohn's disease in the Quebec Founder Population identifies multiple validated disease loci., vol. ”