Artificial intelligence reveal connections among sex, gene methylation, maternal risk factors and disease severity in Autism Spectrum Disorder

Background/Objectives: Increasing evidence points to a contribution of environmental and epigenetic factors in autism spectrum disorder (ASD), but their connections are still largely unexplored. In the present study we used machine learning tools to unravel connections among ASD-related gene methylation levels, maternal ASD risk factors and ASD severity. Methods: The methylation levels of MECP2, OXTR, RELN, BDNF, EN2, BCL2 and HTR1A genes have been assessed in blood DNA samples of 58 ASD children (23 males and 35 females). We then used machine learning approaches (Auto-CM) to connect gene methylation levels with maternal ASD risk factors and with disease severity (ADOS-2 score). Results: Sex differences were observed in DNA methylation levels of the studied genes, with MECP2, HTR1A, and OXTR methylation connected to females, and EN2, BCL2, and RELN methylation connected to males. BDNF methylation was not linked to sex, but rather to maternal risk factors. Maternal prepregnancy BMI, gestational weight gain and living context were among factors linked to disease severity. Conclusion: The present study highlights the power of artificial intelligence tools to unravel connections among different variables in complex disorders, revealing links among maternal risk factors and disease severity or gene methylation levels, as well as sex differences in gene methylation levels that warrant further investigation in ASD.