Towards Prediction and Prioritization of disease genes by the modularity of human phenome-genome assembled network

Abstract

SummaryEmpirical clinical studies on the human interactome and phenome not only illustrates prevalent phenotypic overlap and genetic overlap between diseases, but also reveals a modular organization of the genetic landscape of human disease, provding new opportunities to reduce the complexity in dissecting the phenotype-genotype association. We here introduce a network-module based method towards phenotype-genotype association inference and disease gene identification. This approach incorporates protein-protein interaction network, phenotype similarity network and known phenotype-genotype associations into an assembled network. We then decomposes the resulted network into modules (or communities) wherein we identified and prioritized the disease genes from the candidates within the loci associated with the query disease using a linear regression model and concordance score. For the known phenotype-gene associations in the OMIM database, we used the leave-one-out validation to evaluate the feasibility of our method, and successfully ranked known disease genes at top 1 in 887 out of 1807 cases. Moreover, applying this approach on 850 OMIM loci characterized by an unknown molecular basis, we propose high-probability candidates for 81 genetic diseases.

Structural bounds on the dyadic effect

“…Regarding the applications, when a network and a set of characteristics are given, it is straightforward to compute for each characteristic the point of maximum degeneration D = H = 1, the values D, H and through the use of the four bounds, D min , D max , H min and H max . This approach can be useful in many applications, such as in [19], [20], [21] and [31] where the phase diagram is hard to compute. In such contexts statistical approaches are used to gather information on the correlation between nodes' characteristics and the network topology by looking for the relative distance of the point (D, H) from the point of maximum degeneration.…”

“…The dyadic effect has been considered in order to assess the functional role of nodes within biological networks such as, for instance, in gene-gene interaction in statistical epistasis networks [19], in phenome-genome networks [20] in disease-phenotype network [21] and in protein-protein interaction networks [31] where numerous characteristics are studied to evaluate genetic interactions. Nodes' characteristics are investigated also in interorganizational innovation networks ( [12], [14]) where partnerships agreements of technological transfer among countries are related to innovation indices.…”

“…The literature has attempted to overcome these difficulties by using heuristics or statistical methods. For instance, in [28] it is reported a heuristic method able to identify extremal configurations; in [19], [20], [21], and [31] statistical methods are used to infer the existence of some configurations in computational biology, while in [5] entropy-based measures are used to globally assess the relevance of nodes' characteristics.…”

Endobiogeny: A Global Approach to Systems Biology (Part 1 of 2)

“…New network-based classification methods, such as the human disease network, also known as "diseasome" 54 (Figure 1), have found that disorders should be grouped neither by symptom nor single gene mutations but based on clusters of underlying physiologic dysfunction related to multiple simultaneous genetic polymorphisms and epigenetic changes. [54][55][56][57][58][59][60][61][62] From the clinical perspective, this helps explain why two patients with an "identical" cancer-of the breast, for example-will have different responses to identical chemotherapy regimens. A landmark study of 2000 specimens of breast cancer suggests that it can be subclassified into 10 distinct groups based on various genomic and transcriptomic properties.…”

DiME: A Scalable Disease Module Identification Algorithm with Application to Glioma Progression

“…Though it is capable of yielding biological insight in several case studies (e.g. [4] [5] [6] ), a major drawback of the community detection algorithm is the resolution limit problem [7] [8] which results in huge modules with large numbers of genes (e.g., in [5] ). Such problem is serious in disease module identification since it will inevitably introduce a lot of false disease genes (hence low specificity) and consequently adds difficulties to validation and interpretation.…”