Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2018 Jan;20(1):151-158.
doi: 10.1038/gim.2017.108. Epub 2017 Jul 20.

Knowledge Base and Mini-Expert Platform for the Diagnosis of Inborn Errors of Metabolism

Free PMC article

Knowledge Base and Mini-Expert Platform for the Diagnosis of Inborn Errors of Metabolism

Jessica J Y Lee et al. Genet Med. .
Free PMC article


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.

Conflict of interest statement

The authors declare no conflict of interest.


Figure 1
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
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).

Similar articles

See all similar articles

Cited by 9 articles

See all "Cited by" articles


    1. Tarailo-Graovac M, Shyr C, Ross CJ et al. Exome sequencing and the management of neurometabolic disorders. N Engl J Med 2016;374:2246–2255. - PMC - PubMed
    1. Chong JX, Buckingham KJ, Jhangiani SN et al. The genetic basis of Mendelian phenotypes: discoveries, challenges, and opportunities. Am J Hum Genet 2015;97:199–215. - PMC - PubMed
    1. Survey of the delay in diagnosis for 8 rare diseases in Europe (“EurordisCare2”). Eurodis - Rare Diseases Europe. Accessed 1 November 2016.
    1. Shashi V, McConkie-Rosell A, Rosell B et al. The utility of the traditional medical genetics diagnostic evaluation in the context of next-generation sequencing for undiagnosed genetic disorders. Genet Med 2014;16:176–182. - PubMed
    1. Rare diseases: understanding this public health priority. Eurodis—Rare Diseases Europe. Accessed 1 November 2016.

Publication types