RDAD: Phenotype-based Rare Disease Auxiliary Diagnosis System

Rare disease is a disease with a precious low incidence and most of them are chronic.The rare disease associated phenotypes are usually described by a set of clinical medical terms.Rare diseases always have a wide range of complex and diverse phenotypes,however,clinicians always lack of either awareness of rare diseases or clinical experiences,which makes patients with rare diseases often not be accurately diagnosed and effectively treated on time.To integrate multi-level biomedical resources and multiple classifiers,we applied the phenotypic TF-IDF-Hierarchy information content on the phenotype semantic hierarchy of Human Phenotype Ontology(HPO),and then built the phenotypic TFIDF-Hierarchy information content based rare disease similarity model(PICS),the phenotype-gene association based rare disease similarity model(PGAS)and the curated feature phenotype spatial vector based rare disease machine learning prediction model(CPML),as well as the curated and text mined feature phenotype spatial vector based rare disease machine learning prediction model(APML).We validated our four diagnostic models with the real medical records from RAMEDIS.The result shows that all the above four models have high diagnostic precision(?98%)with acceptable recall(?67%),and the CPML model achieves the highest precision(?99%)and the highest recall(?95%),and is also the diagnostic model that we recommend.To promote effective diagnosis for rare disease in clinical application,we developed the phenotype-based Rare Disease Auxiliary Diagnosis system(RDAD)to assist clinicians to diagnose rare diseases with the above four diagnostic models,user can search our system with patients' phenotypes.The RDAD web server can be accessed through the link http://www.unimd.org/RDAD/.