AI-Driven Insights into Neurobiological Mechanisms: Machine Learning Applications in Dyslexia and Autism Spectrum Disorder

Dyslexia and autism spectrum disorder (ASD) are neurodevelopmental conditions characterized by distinctive neurobiological patterns, including alterations in synaptic connectivity, neuroimmune regulation, and electrophysiological activity. Advances in artificial intelligence (AI) and machine learning (ML) offer powerful tools to detect and analyze these complex patterns, potentially improving early diagnosis and understanding of underlying mechanisms. This paper reviews how AI-driven analysis of neurobiological data—particularly electroencephalography (EEG) biomarkers—yields insights into dyslexia and ASD. Dyslexia is associated with aberrant neural connectivity and increased slow-wave EEG activity, linked to excessive synaptic pruning and neuroinflammation. ASD features atypical connectivity (often hyperconnectivity) and a "U-shaped" EEG power profile reflecting insufficient pruning and chronic neuroimmune activation. Machine learning algorithms can classify individuals with dyslexia or ASD based on such EEG biomarkers with high accuracy, as demonstrated by recent studies achieving ~95–99% classification performance. These models have identified key features—for example, elevated theta and reduced beta1 band power in dyslexia—that correlate with cognitive deficits and microglial overactivation. We discuss how AI facilitates multi-modal integration of synaptic, immune, and EEG data, helping unravel the neurobiological substrates of dyslexia and ASD. While highly promising, ML-based diagnostic tools face limitations (e.g., false negatives due to overlapping patterns) and require careful validation. Nonetheless, AI-driven analyses provide a novel lens on brain connectivity and neuroimmune dynamics in developmental disorders, potentially guiding personalized interventions and future research.