Knowledge base and mini-expert platform for the diagnosis of inborn errors of metabolism

Abstract

PurposeRecognizing individuals with inherited diseases can be difficult because signs and symptoms often overlap those of common medical conditions. Focusing on inborn errors of metabolism (IEMs), we present a method that brings the knowledge of highly specialized experts to professionals involved in early diagnoses. We introduce IEMbase, an online expert-curated IEM knowledge base combined with a prototype diagnosis support (mini-expert) system.MethodsDisease-characterizing profiles of specific biochemical markers and clinical symptoms were extracted from an expert-compiled IEM database. A mini-expert system algorithm was developed using cosine similarity and semantic similarity. The system was evaluated using 190 retrospective cases with established diagnoses, collected from 15 different metabolic centers.ResultsIEMbase provides 530 well-defined IEM profiles and matches a user-provided phenotypic profile to a list of candidate diagnoses/genes. The mini-expert system matched 62% of the retrospective cases to the exact diagnosis and 86% of the cases to a correct diagnosis within the top five candidates. The use of biochemical features in IEM annotations resulted in 41% more exact phenotype matches than clinical features alone.ConclusionIEMbase offers a central IEM knowledge repository for many genetic diagnostic centers and clinical communities seeking support in the diagnosis of IEMs.

Figure 1

Mini-expert algorithm flowchart. Users enter a list of biochemical/clinical phenotypes into IEMbase’s mini-expert system. The system’s phenotype-matching algorithm first divides the input list into biochemical and clinical categories. The algorithm then ranks the disorders in IEMbase by comparing the biochemical profile of each disorder against the input biochemical profile, using cosine similarity. Subsequently, the algorithm breaks ties in the ranked list by comparing the clinical profiles, using semantic similarity.

Figure 2

Mini-expert system performance using only biochemical/clinical information. The system performance when using only biochemical phenotypes was compared with that when using only clinical phenotypes of 172 retrospective cases. Percentage success N measures % of cases whose actual diagnoses ranked within the top N ranks. The system performance when using only biochemical phenotypes was significantly better than that when using only clinical phenotypes (P < 2.2e-16; Mann-Whitney-U).