Authors (including presenting author) :
Wong KC(1), Law CY(1), Lam CW(1)(2)
Affiliation :
(1) Division of Chemical Pathology, Department of Pathology, Queen Mary Hospital (2) Department of Pathology, The University of Hong Kong
Introduction :
Urine Organic Acids (UOA) analysis is an essential investigation for workup of Inborn Errors of Metabolism (IEM). Similar to many centers, our laboratory employed a gas chromatography-mass spectrometry (GC-MS) system for UOA analysis which covers >100 IEM-related metabolites (biomarkers) in a single run. Nevertheless, there are several possible pitfalls, including low level metabolites, co-elution, age-dependent variation of metabolite levels, etc. Thus, the interpretation is a labor-intensive process and takes considerable time from Chemical Pathologists.
Objectives :
An effective automatic solution was established to address the above pitfalls and streamline the UOA reporting process.
Methodology :
An in-house database composed of 94 key metabolites was built up using existing 1,642 sets of UOA GC-MS data and partitioned according to different age groups. Positive identification of metabolites was defined according to their retention times (RT) and electron ionization spectrum in GC-MS. The 95th percentile for each compound in each age group was used as a cutoff to define abnormally high OAs, which were automatically highlighted in spreadsheet format for review by Chemical Pathologists. The diagnostic performance of the automatic solution was tested using cases from External Quality Assurance Programs (EQAP).
Result & Outcome :
Among the 23 positive and 10 negative cases from EQAP from 2017 to 2019, 22 positive (95.7%) and 10 negative (100%) cases were correctly identified, with a diagnostic accuracy of 97.0%. The false negative case is mevalonic kinase deficiency in a clinically well subject, as the pathognomonic biomarker mevalonolactone was expected to be very low, whereas in the urine of this same subject during symptomatic phase, mevalonolactone was correctly identified. Our algorithm also allows (1) graphical display of individual UOA levels and comparison with controls according to different age groups, (2) calculation of ratios of metabolites which is useful in interpreting low level but clinically significant metabolite (e.g. vanillactate), (3) pathway analysis by a holistic correlation analysis of all studied OAs, and (4) continual database enrichment.
Conclusion: The automatic algorithm is a new and effective tool for rapid interpretation of UOA results with high diagnostic accuracy. The algorithm can be used as a comparator for automatic interpretation of UOA results based on machine learning.
