Immune-related DNA methylation signatures, age prediction, and cross-tissue assessment via biologically informed neural networks.

Abstract

Immune-related DNA methylation patterns may provide interpretable markers of biological ageing and disease relevance. This exploratory study integrated four Illumina HumanMethylation450 and MethylationEPIC datasets comprising paediatric peripheral-blood leukocytes (GSE36054; 192 retained after quality control), infant buccal cells (n = 542), neuronal nuclei (n = 58), and peripheral blood from a polybrominated-biphenyl exposure cohort (n = 659). The paediatric peripheral-blood dataset served as the primary modelling cohort, with infant buccal cells providing an independent early-life cross-tissue reference. Processing included cohort-specific quality control and normalisation, batch adjustment, empirical-Bayes differential methylation analysis, pathway enrichment, repeated 10-fold cross-validation, biologically informed neural-network modelling, feature attribution, cross-tissue assessment, gene-ablation analysis, disease mapping, and two-sample Mendelian randomisation. IL6, CXCR4, CDK11 and CREBBP showed the strongest coherent methylation signals and stable predictive contributions, whereas BCLAF1 contributed less consistently. These loci mapped to cytokine-receptor interaction, chemokine signalling, G2/M checkpoint regulation, histone acetylation, and apoptosis-related pathways. The neural network achieved a cross-validated mean absolute error of 1.87 years and R2 of 0.781; removing the polybrominated-biphenyl-associated feature filter yielded comparable performance (MAE 1.91 years; R2 = 0.772). Prioritised features also showed cross-tissue consistency in infant buccal cells. Mendelian randomisation identified genetically supported associations of IL6 methylation with preterm birth and lower mean birth weight and of CXCR4 methylation with preterm birth after multiplicity correction. These associations do not establish that methylation itself causes the outcomes. The findings prioritise immune-related methylation features across paediatric blood and infant buccal contexts within an interpretable framework for age-related methylation and outcome-oriented molecular investigation.