Autism Spectrum Disorder (ASD) is a heterogeneous neurodevelopmental disorder that has no known etiology, but has many known risk factors, including preterm birth and environmental exposures such as BPA. Both placental inflammation and morphological changes have been associated with the development of ASD later in life. Due to the heterogeneous nature of ASD and the wide range of potential causative genes, this study used machine-learning methods to identify candidate predictive genes for ASD at age 10, based on placental CpG methylation and mRNA transcript data. Using data from the Extremely Low Gestational Age Newborns (ELGAN) cohort, the Elastic Net models found 15 candidate predictive genes from the CpG model and 56 candidate predictive genes from the mRNA model. These candidate predictors included 3 from the CpG model and 16 from the mRNA model that had previously been associated with ASD, leaving 52 novel candidate predictive genes. A pathway enrichment analysis using STRING showed significant enrichment and identified key pathways, including the Ubiquitin-Proteasome system (UPS) involving CUL2, PSMD4, PSMD7, and RPS16, and diacylglycerol metabolism involving PLCE1 and DGKE. The UPS elucidated a potential mechanism by which placental inflammation may impact ASD outcomes later in life. Furthermore, BPA was identified in the Comparative Toxicogenomics Database as a toxicant of interest for perturbing the UPS system.
Arjun Suresh (Wed,) studied this question.