Artificial intelligence analysis of newborn leucocyte epigenomic markers for the prediction of autism.

A great diversity of factors contribute to the pathogenesis of autism and autism spectrum disorder (ASD). Early detection is known to correlate with improved long term outcomes. There is therefore intense scientific interest in the pathogenesis of and early prediction of autism. Recent reports suggest that epigenetic alterations may play a vital role in disease pathophysiology. We conducted an epigenome-wide analysis of newborn leucocyte (blood spot) DNA in autism as defined at the time of sample collection. Our goal was to investigate the epigenetic basis of autism and identification of early biomarkers for disease prediction. Infinium HumanMethylation450 BeadChip assay was performed to measure DNA methylation level in 14 autism cases and 10 controls. The accuracy of cytosine methylation for autism detection using six different Machine Learning/Artificial Intelligence (AI) approaches including Deep-Learning (DL) was determined. Ingenuity Pathway Analysis (IPA) was further used to interrogate autism pathogenesis by identifying over-represented biological pathways. We found highly significant dysregulation of CpG methylation in 230 loci (249 genes). DL yielded an AUC (95% CI) = 1.00 (0.80-1.00) with 97.5% sensitivity and 100.0% specificity for autism detection. Epigenetic dysregulation was identified in several important candidate genes including some previously linked to autism development e.g.: EIF4E, FYN, SHANK1, VIM, LMX1B, GABRB1, SDHAP3 and PACS2. We observed significant enrichment of molecular pathways involved in neuroinflammation signaling, synaptic long term potentiation, serotonin degradation, mTOR signaling and signaling by Rho-Family GTPases. Our findings suggest significant epigenetic role in autism development and epigenetic markers appeared highly accurate for newborn prediction.