Fisher discriminant analysis for classification of autism spectrum disorders based on folate-related metabolism markers.

Autism spectrum disorders (ASDs) are neurodevelopmental disorders with an increasing prevalence but lack reliable biomarkers for early diagnosis. The present study investigated 13 serological metabolites and 2 genetic variants related to folate metabolism in a total of 89 ASD cases and 89 matched controls. Fisher discriminant analysis was used to establish the classification model to recognize ASD cases and controls. Ten metabolites were significantly different between the groups, of which six metabolites were used as predictors to determine the discriminant prediction model: vitamin B12, 5-methylene-tetrahydrofolate, methonine, the ratio of S-adenosylmethionine/S-adenosylhomocysteine, methionine synthase and transcobalamin II. The model had statistical significance (lambda=0.520, χ2=113.103, df=6, P<.001) and correctly identified 84.3% of ASD and normal cohorts. The area under the receiver operating characteristic curve was 0.913, with a sensitivity of 86.5% and a specificity of 85.4%. Overall, the results indicated that folate-related metabolism contributed to predisposition of ASD and the combined detection of folate-related metabolism biomarkers could be effective in distinguishing ASD from healthy controls, and provide new insights for the early diagnosis of ASD in the future.