Introduction: The etiologies of Autism Spectrum Disorder (ASD) are complex, with evidence suggesting that genetics, maternal in utero exposures, autoantibodies, and the early-life exposome contribute to disease progression. We identify multiple converging factors contributing to ASD, highlighting a significant association between sepsis at birth and later ASD diagnosis through complementary methodologies. Methods: Using electronic health records (EHR) from the Corewell Health system (2017–2024), we linked ASD diagnoses to maternal pregnancy complications and neonatal diagnoses. Genetic and environmental risk models were built using published stem cell-derived brain organoid exposure studies (PMID: 33657557), large-scale ASD gene datasets integrated through AI-driven analytics, and the Comparative Toxicogenomics Database. To capture lived experience and care complexity, 100 caregiver and clinician narratives were created through AI-simulated journeys using converged online chat forum content within our Baymax (Behavioral and ArtificiallY-Modeled Analysis eXchange) acute-to-chronic risk modeling framework. Results: Both the Baymax model and our EHR analysis of 932 ASD cases revealed overlapping clinical risk factors. Enriched neonatal conditions from the EHR included probable sepsis (68 cases, ~7× expected), transient tachypnea (55 cases, ~3–5×), neonatal hypoglycemia (57 cases, ~2–3×), and in utero drug exposure (49 cases, ~2×). Baymax modeling identified intrauterine growth restriction (12%), sepsis (12%), preeclampsia (9%), and gestational diabetes (8%) as frequent contributors. ASD exposome-gene mapping further highlighted risk genes responsive to inflammatory and infectious stress, supporting a cytokine–neurodevelopmental axis that links early sepsis to ASD pathogenesis. Conclusions: This work demonstrates a multi-tool framework integrating clinical data, experimental biology, environmental databases, and AI-enhanced storytelling to identify sepsis at birth as a candidate contributor to ASD risk. These insights support the development of future intervention strategies to mitigate neurodevelopmental complications among at-risk neonates.
Prokop et al. (Sun,) studied this question.