The clinical spectrum of Alzheimer's disease (AD) ranges dynamically from asymptomatic and mild cognitive impairment (MCI) to mild, moderate, or severe AD. Although a few disease-modifying treatments, such as lecanemab and donanemab, have been developed, current therapies can only delay disease progression rather than halt it entirely. Therefore, the early detection of MCI and the identification of MCI patients at high risk of progression to AD remain urgent unmet needs in the super-aged era. This study utilized transcriptomics data from cognitively unimpaired (CU) individuals, MCI, and AD patients in the Alzheimer's Disease Neuroimaging Initiative (ADNI) cohort and leveraged machine learning models to identify biomarkers that differentiate MCI from CU and also distinguish AD from MCI individuals. Furthermore, Cox proportional hazards analysis was conducted to identify biomarkers predictive of the progression from MCI to AD. Our machine learning models identified a unique set of gene expression profiles capable of achieving an area under the curve (AUC) of 0.98 in distinguishing those with MCI from CU individuals. A subset of these biomarkers was also found to be significantly associated with the risk of progression from MCI to AD. A linear mixed model demonstrated that plasma tau phosphorylated at threonine 181 (pTau181) and neurofilament light chain (NFL) exhibit the prognostic value in predicting cognitive decline longitudinally. These findings underscore the potential of integrating machine learning (ML) with transcriptomic profiling in the early detection and prognostication of AD. This integrated approach could facilitate the development of novel diagnostic tools and therapeutic strategies aimed at delaying or preventing the onset of AD in at-risk individuals. Future studies should focus on validating these biomarkers in larger, independent cohorts and further investigating their roles in AD pathogenesis.
Keywords: Alzheimer’s disease (AD); MCI-to-AD conversion; RNA sequencing; biomarkers; gene expression; machine learning; mild cognitive impairment (MCI); transcriptomics.