Network based approach for discovering novel gene-phenotypic association and disease co morbidities using ontological data

Abstract Advancements in bio-techniques have accelerated the generation of genomic and proteomic data. High throughput experiments generate huge volume of biological data that lead to cutting edge researches. These data may be stored in hierarchical structures known as ontologies. Ontologies serve as rich knowledge sources for biological information mining. Annotations from Gene Ontology (GO), Human Phenotype Ontology (HPO) and Drug Ontology are extensively used for biological studies including disease causing gene prediction, disease comorbidity analysis and drug designing purposes. Several mechanisms exist for extracting semantic similarity among ontological terms. Graph based approaches are mostly used for analyzing the semantic similarity between ontological terms. In this paper, a novel method has been proposed for predicting disease causing genes and disease comorbidities using GO and HPO data. Weighted association rules obtained from Gene Ontology and Human Phenotype Ontology are used for constructing weighted gene network and weighted phenotype network. These networks are then connected using known gene-phenotypic relationships. Modified random walk restart algorithm is performed on this network for extracting novel disease gene correlations and disease comorbidities. This method outperformed the existing methods that depend on similarity measurements.