Automated screening of metabolic disorders using pattern recognition of GC-MS full scan spectra from urine organic acids

Analysis of organic acids in urine is a valuable tool in the diagnosis of the inborn errors of metabolism known as organic acidurias. This test is commonly ordered in newborns with symptoms such as lethargy, failure to thrive, hepatic failure, and suspected familial disorders. A drawback of published methods is the overwhelming amount of data to examine for each patient, prior to the final laboratory report. Physicians will wait as long as two weeks for these time critical results. The goal of this research was to develop and export system to automate the process of screening for metabolic disorders of urine organic acids. The Xaminer® pattern recognition software (ThermoFinnigan, San Jose, CA) was adapted to predict and identify patterns of urine organic acid disorders. The gas chromatography-mass spectrometry (GC-MS) full scan spectra of organic acids were used to build the pattern match library and train the software to recognized methylmalonic aciduria (MMA) and associated vitamin B12 deficiency, as well as, a subset of fatty acid oxidation defects (FAOD), including medium chain acyl-CoA dehydrogenase (MCAD) deficiency. Patient data files were de-identified and reprocessed using the expert system. The expert system results were compared to the original laboratory findings. From a total of 2573 samples, the original laboratory findings were 20 positives for MMA and 29 positives for FAOD. The Xaminer software identified 17 of the 20 MMA positives, plus 4 additional candidate samples that matched the search pattern criteria. The software found 26 of the 29 FAOD positives. Five additional samplers found to be candidates for FAOD. Software analysis time averaged less than 10 seconds per sample. This expert system can use pattern recognition of full scan GC-MS data to aid in patient screening for MMA and fatty acid oxidation disorders. The performance of Xaminer shows promise for refining or expanding the reference library to include other metabolic disorders as well.