Alzheimer's disease (AD) is a heterogeneous neurodegenerative condition. This study identifies clinically relevant new molecular subtypes of the early and late mild cognitive impairment (EMCI and LMCI) stages of AD from 401 patients in the ADNI consortium. Metabolomics and peripheral blood mononuclear cell (PBMC) transcriptomics data are integrated using Similarity Network Fusion (SNF), resulting in two molecular subtypes within EMCI (EMCI-1 and EMCI-2) and LMCI (LMCI-1 and LMCI-2), respectively. Metabolomics, rather than gene expression profiling, is the major contributor to subtyping. Subtype-specific differences in metabolite levels and gene expression correlate with AD pathophysiology, supported by cognitive scores, pseudo-trajectory analysis, and longitudinal dementia diagnoses. The new molecular subtypes are further validated through the EFIGA dataset, and present a clear trend of increase in protein biomarkers including p-Tau181, p-Tau217, Aβ40, and Aβ42. These refined subtypes represent a step toward more personalized therapeutic strategies targeting early disease stages before AD diagnosis.
Keywords: Alzheimer’s disease; Mild cognitive impairment; Multi-omics integration; Similarity Network Fusion; metabolomics; transcriptomics.