Decoding Pathogenic Mutational Landscapes in Alzheimer's Disease Through Integrated Transcriptomics

Alzheimer's disease (AD) is increasingly understood as a disorder driven not only by amyloid and tau pathology but also fundamentally shaped by underlying genetic mutations. By integrating multiple AD gene expression datasets with machine learning approaches-including random forest, XGBoost, LASSO, and SVM-we identified 172 differentially expressed genes, with TUBB2A, RTN4, and YWHAZ emerging as top mutation-associated hub genes. Critically, TUBB2A not only exhibited strong diagnostic potential (AUC = 0.822) but also harbored somatic mutations in our patient cohort, directly linking mutational events to disease manifestation. Unsupervised clustering revealed two distinct AD subtypes: one marked by widespread early gene overexpression and another (Cluster 2) dominated by endoplasmic reticulum stress-likely reflecting divergent mutational landscapes. Pseudotemporal trajectory analysis demonstrated a continuous progression from normal samples to Cluster 2, suggesting that a pivotal mutational event may initiate this transition and accelerate disease progression. These findings underscore the central role of somatic and germline mutations-particularly in TUBB2A-in AD pathogenesis. Our study strongly supports a paradigm shift toward mutation-centric biomarker development and advocates for SNP-based strategies to enable early diagnosis and personalized therapeutic interventions tailored to individual mutational profiles.