Identification of Immune Infiltration and Iron Metabolism-Related Subgroups in Autism Spectrum Disorder

Abstract

Autism spectrum disorder (ASD) is a prevalent neurodevelopmental disorder with a broad spectrum of symptoms and prognoses. Effective therapy requires understanding this variability. ASD children's cognitive and immunological development may depend on iron homoeostasis. This study employs a machine learning model that focuses on iron metabolism hub genes to identify ASD subgroups and describe immune infiltration patterns. A total of 97 control and 148 ASD samples were obtained from the GEO database. Differentially expressed genes (DEGs) and an iron metabolism gene collection achieved the intersection of 25 genes. Unsupervised cluster analysis determined molecular subgroups in individuals with ASD based on 25 genes related to iron metabolism. We assessed gene ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment, gene set variation analysis (GSVA), and immune infiltration analysis to compare iron metabolism subtype effects. We employed machine learning to identify subtype-predicting hub genes and utilized both training and validation sets to assess gene subtype prediction accuracy. ASD can be classified into two iron-metabolizing molecular clusters. Metabolic enrichment pathways differed between clusters. Immune infiltration showed that clusters differed immunologically. Cluster 2 had better immunological scores and more immune cells, indicating a stronger immune response. Machine learning screening identified SELENBP1 and CAND1 as important genes in ASD's iron metabolism signaling pathway. These genes express in the brain and have AUC values over 0.8, implying significant predictive power. The present study introduces iron metabolism signaling pathway indicators to predict ASD subtypes. ASD is linked to immune cell infiltration and iron metabolism disorders.

Keywords: Autism spectrum disorder; Immune infiltration; Iron metabolism; Machine learning; Molecular clusters.