Machine Learning-Based Identification of Immune Inflammation-Related Genes as Shared Potential Diagnostic Biomarkers in Autism Spectrum Disorder and Atopic Dermatitis

Background: ASD is a class of neurodevelopmental disorders with onset in early childhood, whereas AD is a common chronic inflammatory skin disease. An increasing number of studies suggest that immune dysregulation and inflammatory responses play important roles in the onset and progression of both conditions; however, their shared molecular mechanisms remain unclear. Methods: First, ASD-related and AD-related datasets were obtained from the GEO database. After removal of batch effects, the common DEGs between the two diseases were identified. Subsequently, 107 machine learning-based model configurations were employed to screen for key genes. Functional enrichment analyses and PPI network construction were performed to systematically explore their potential functions. Finally, the CIBERSORT was applied to analyze immune cell infiltration and to assess the correlation between hub gene expression and immune cell infiltration. Results: 164 common genes between ASD and AD were identified. GO and KEGG enrichment analyses revealed that these shared differentially expressed genes were mainly enriched in pathways related to immune regulation and inflammatory responses, suggesting that immuno-inflammatory processes may constitute an important biological basis linking ASD and AD. Further screening and validation using machine learning identified BEX4, BIN2, BNIP3L, CCNO, JAK2, SLC39A7, and WASF3 as hub genes serving as common potential biomarkers for both diseases. Among them, BIN2, SLC39A7, and JAK2 may represent key shared genes and demonstrated good diagnostic value in ROC curve and nomogram analyses. In addition, immune infiltration analysis indicated that these key genes were significantly correlated with the infiltration levels of multiple immune cell types, further supporting their potential roles in immune regulation. Conclusions: This study reveals potential shared immuno-inflammatory molecular mechanisms between ASD and AD. Genes screened based on 107 machine learning models were verified as potential diagnostic biomarkers for both diseases after integrated analysis, providing a theoretical basis for further investigation of their immune-related pathogenesis and early clinical diagnosis.