PhenomeNET: a whole-phenome approach to disease gene discovery

Abstract

Phenotypes are investigated in model organisms to understand and reveal the molecular mechanisms underlying disease. Phenotype ontologies were developed to capture and compare phenotypes within the context of a single species. Recently, these ontologies were augmented with formal class definitions that may be utilized to integrate phenotypic data and enable the direct comparison of phenotypes between different species. We have developed a method to transform phenotype ontologies into a formal representation, combine phenotype ontologies with anatomy ontologies, and apply a measure of semantic similarity to construct the PhenomeNET cross-species phenotype network. We demonstrate that PhenomeNET can identify orthologous genes, genes involved in the same pathway and gene-disease associations through the comparison of mutant phenotypes. We provide evidence that the Adam19 and Fgf15 genes in mice are involved in the tetralogy of Fallot, and, using zebrafish phenotypes, propose the hypothesis that the mammalian homologs of Cx36.7 and Nkx2.5 lie in a pathway controlling cardiac morphogenesis and electrical conductivity which, when defective, cause the tetralogy of Fallot phenotype. Our method implements a whole-phenome approach toward disease gene discovery and can be applied to prioritize genes for rare and orphan diseases for which the molecular basis is unknown.

Figure 1.

Overview over ontology-based data analysis. First, the ontologies have to be formalized before their consistency can be verified. If contradictory axioms are identified, they must be removed. Using the ontology, biological data is represented within the same model so that the biological questions across the data can be asked in flexible ways. If necessary, statistical approaches are applied to complete missing information and results can then be inferred over the combined representation.

Figure 2.

Illustration of assertions and inferences about the class Matted coat. Blue-colored shapes represent qualities, gray-colored shapes represent anatomical entities and green-colored shapes represent phenotypes. Dashed lines represent inferred associations.

Figure 3.

ROC curves for predicting disease, participation in a common pathway and orthology using PhenomeNET. The ROC curves for pathway and orthology predictions are obtained by comparison with KEGG, while the gene-disease predictions are derived from OMIM and the annotated disease models in the MGI. AUC for pathways is 0.59, for orthology 0.62 and for disease 0.68.