Screening biomarkers for autism spectrum disorder using plasma proteomics combined with machine learning methods.

Background And Aims: Autism spectrum disorder (ASD) is a common neurodevelopmental disorder in children. Early intervention is effective. Investigation of novel blood biomarkers of ASD facilitates early detection and intervention.

Materials And Methods: Sequential window acquisition of all theoretical spectra-mass spectrometry (SWATH-MS)-based proteomics technology and 30 DSM-V defined ASD cases versus age- and sex-matched controls were initially evaluated, and candidate biomarkers were screened using machine learning methods. Candidate biomarkers were validated by targeted proteomics multiple reaction monitoring (MRM) analysis using an independent group of 30 ASD cases vs. controls.

Results: Fifty-one differentially expressed proteins (DEPs) were identified by SWATH analysis. They were associated with the immune response, complements and coagulation cascade pathways, and apolipoprotein-related metabolic pathways. Machine learning analysis screened 10 proteins as biomarker combinations (TFRC, PPBP, APCS, ALDH1A1, CD5L, SPARC, FGG, SHBG, S100A9, and PF4V1). In the MRM analysis, four proteins (PPBP, APCS, FGG, and PF4V1) were significantly different between the groups, and their combination as a screening indicator showed high potential (AUC = 0.8087, 95 % confidence interval 0.6904-0.9252, p < 0.0001).

Conclusions: Our study provides data that suggests that a few plasma proteins have potential use in screening for ASD.