Modular-based Mapping Between The Human Phenotype And The Mammalian Phenotype Provides A New Approach To Predict The Disease Pathogenic Gene

Mapping human phenotypes to model organism phenotypes plays an important role in the study of disease mechanisms.Mouse and other mammalians have been used as models to decipher the underlying mechanisms of various human phenotypes,including diseases.Mapping between human phenotypes and mammalian phenotypes has been regarded as an effective approach to interpret the potential functions of human genes.Currently,the mapping between human phenotypes and mammalian phenotypes are generally based on two approaches – connecting human phenotype to mammalian phenotype based on their associations through semantic similarities between phenotypic terms from HPO(Human Phenotype Ontology)and MPO(Mammalian Phenotype Ontology)and the ontologies' taxonomic structures,while the second method designing mapping algorithm based on homologous genes.However,given that the different representation models of mutant mouse phenotype and human disease abnormalities as well as their inconsistency of hierarchical structures,the results of the two methods have shown a very limited result and can only cover a little phenotype terms.Here,we proposed a modular recognition method which can recognize significant module-module correlations between HPO and MPO phenotypes.The phenotype modules were developed by detecting phenotype communities which showed a steady underlying mechanism interactions in the gene-based phenotype network.The connections between human phenotype communities and mammalian phenotype communities were built by linking their corresponding homologous genes based on gene-based phenotype network.As a result,7805 human phenotype communities and 5447 mammalian phenotype communities are built on 13537 human phenotype terms and 8810 mammalian phenotype terms.Through mapping human phenotype communities to mammalian phenotype communities based on the homologous genes and thenanalyzing the phenotype terms in the communities,we obtained 3223 human phentype community and mammalian phenotype community pairs.Based on the mapping results,we then constructed a model to predict causative genes for those disease phenotypes.We also adopted statistical methods and manual analysis to validate the predictive results,which showed that the modularized approach could generate reliable mapping and predict candidate genes.Taken together,our module-based mapping approach between HPO and MPO,as well as analyses on those modules provide an important resource for the underlying mechanism investigation of human phenotypes.