Machine learning-based computational prediction of sirtuin-NF-κB regulatory axis driving inflammatory cell death in Alzheimer’s disease

This is the authors' abstract. We don't add key points for this paper.

Objective To identify molecular regulators linking neuroinflammation, epigenetic regulation, and inflammatory programmed cell death in Alzheimer’s disease (AD) and evaluate their diagnostic relevance. Methods An integrative computational framework was applied to the cortical transcriptomic dataset GSE15222 (176 AD cases; 187 controls), incorporating differential expression, functional enrichment, protein-protein interaction (PPI) analysis, five machine-learning (ML) classifiers, explainable artificial intelligence (XAI), structural-disorder prediction, and molecular docking. Candidate biomarkers were externally evaluated using the independent blood-based GSE63060 cohort. Results GSE15222 identified 2285 differentially expressed mRNAs (|log2FC= > 0.5, P < 0.05), including 19 genes overlapping pyroptosis- and necroptosis-related signatures. Enrichment analysis identified NF-κB, inflammasome, cytokine, and cell-death pathways. XGBoost (eXtreme Gradient Boosting) achieved the best performance (ROC-AUC: 0.889; accuracy: 0.835; F1-score: 0.830; Matthews correlation coefficient: 0.671). Cross-validation showed stable prioritisation of NFKB1 (SHAP, 0.184 ± 0.021; top-10 frequency, 100%), NFKBIA (0.161 ± 0.019; 100%), and XIAP (0.148 ± 0.018; 100%). In GSE63060, NFKB1(Nuclear Factor Kappa B subunit 1), NFKBIA (NF- κB inhibitor alpha), and XIAP (X-linked Inhibitor of Apoptosis Protein) achieved ROC-AUCs of 0.822, 0.759, and 0.747, respectively; their expression differences were significant ( P < 2.22 × 10⁻¹⁶ for each), while NLRP3 (NLR family pyrin domain containing 3) showed P = 2.7 × 10⁻⁷. Structural analysis yielded disorder scores of 0.6043, 0.0231, and 0.6089 for NFKBIA-K67, XIAP-K493, and NFKB1-K967, respectively. Docking predicted interaction energies of −786.3, −1052.4, and −1158.1 kcal/mol for NFKBIA-SIRT1, XIAP-SIRT6, and NFKB1-SIRT6, respectively. Conclusion NFKB1, NFKBIA, and XIAP represent robust candidate biomarkers and support a potential sirtuin-NF-κB-XIAP regulatory axis linking epigenetic regulation with inflammatory cell death in AD. The predicted molecular interactions remain computational hypotheses requiring experimental validation.