Diagnosis Of Genetic Disorders Based On The Human Phenotype Ontology

The rapid development of sequencing technologies promote the generation of large amount of genotype data,making it possible for patient’s sequencing data to be used in clinical diagnosis of genetic disorders.However,in real clinical situation,diagnosis of genetic disorders still mainly focuses on phenotype information,which needs large training and rich experience.The establishment of Human Phenotype Ontology provides a new thought to make phenotype based diagnosis from the point of computation.Human Phenotype Ontology is a structured,controlled and standardized vocabulary for systematically describing the semantic relationship between phenotypes and constructing a perfect knowledge database.With the maturity of Human Phenotype Ontology and the enrichment of annotations,it will be of great help to the phenotype based diagnosis of genetic disorders.To make diagnosis of genetic disorders using Human Phenotype Ontology,the popular way is to infer the underlying genetic disorder form the calculation of semantic similarities between patient’s phenotypes and different genetic disorders.Inspired from Jaccard index,true-path rule and One-sided search,in this article,we proposed a new method based on relative importance to calculate the semantic similarities from patient’s phenotypes to different genetic disorders.Through the comparison of predication accuracy on different datasets(a dataset without noise and imprecision,a dataset with noise,a dataset with imprecision,a dataset with both noise and imprecision),it can be concluded that the proposed method is superior or at least equal to seven common semantic similarity measures,implying the advantage and stability of the proposed method.Under the most difficult circumstance,the proposed method achieved the predication accuracy(top 1)of 33.6% in the dataset with both noise and imprecision,17% larger than Resnik measure.