International Neuropsychiatric Disease Journal

Background: Autism Spectrum Disorder (ASD) is a neurodevelopmental disorder with a multifactorial etiology, with a male-to-female predominance of 4:1. Currently, clinical diagnosis is late, missing the window of therapeutic neuroplasticity for timely intervention. Salivary microRNAs (miRNAs) are stable epigenetic regulators that offer a non-invasive collection method. This study aimed to evaluate the evidence on the diagnostic accuracy of salivary miRNAs in ASD and polyomics integration. 

Methods: A systematic review was conducted under PRISMA 2020 guidelines using PubMed and Google Scholar between 2016 and 2025. Observational case-control studies in the pediatric population with a validated ASD diagnosis versus controls were selected. Quality and risk of bias were assessed using the Newcastle-Ottawa Scale (NOS). A total of 657 records were identified, and after screening and eligibility assessment, six studies were included in the qualitative synthesis.

Results: We included six studies with 1,149 participants, 73.2% male and 26.8% female. The predictive models showed high diagnostic performance, a panel of 5 miRNAs achieved an area under the ROC curve (AUC) of 0.952, with sensitivity of 90.32% and specificity of 90%, while the polyomic approach integrating human and microbial RNA reached an AUC value of 0.88 (95% CI:  0.86-0.88), with sensitivity of 82% and specificity of 88%. Individually, miR-451a exhibited significant downregulation in patients with ASD (FC = -3.58; P < 0.0001). Alterations in axon guidance, circadian rhythms, and synaptic signaling pathways were confirmed. A negative correlation (r = -0.30) was found between miR-141-3p and the genus Tannerella.

Conclusion: Salivary miRNA panels demonstrate promising diagnostic discrimination in case-control settings, overcoming subjective behavioral tool limitations, but prospective screening validation remains needed. Non-invasive collection facilitates timely detection within the neuroplasticity window, where early intervention is most effective. Furthermore, integrating host and microbial transcriptomic data supports a multifactorial ASD pathophysiology.