Multi-Omics Modelling of Plasma pTau181, GFAP, and Metabolic Features Enables Risk Stratification in Prodromal Alzheimer’s Disease
Discuss this preprint
Start a discussion What are Sciety discussions?Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Background
Mild cognitive impairment (MCI) is a heterogeneous state between normal ageing and dementia, often considered prodromal to Alzheimer’s disease (AD). Progression is variable, and distinguishing stable from progressive MCI remains difficult, particularly in the presence of mixed neuropathology. Blood biomarkers such as phosphorylated tau181 (pTau181), glial fibrillary acidic protein (GFAP), and neurofilament light chain (NfL) demonstrate prognostic value in established AD, but limited performance for prognosticating progression from MCI.
Methods
Blood protein biomarkers (pTau181, GFAP, NfL) were integrated with NMR- and LC-MS-derived metabolomic features. In a deeply phenotyped MCI cohort (VITACOG; n =68) with two-year MRI follow-up, cross-validated logistic regression identified discriminative multi-analyte panels to distinguish stable from progressive MCI. Disease progression was defined by worsening cortical atrophy, measured via annualised brain volume loss. Generalisability was tested in a larger community-based cohort from UK Biobank ( n =223) and two Oxford Project to Investigate Memory and Ageing (OPTIMA) subsets with histopathological diagnosis ( n =61, n =37).
Results
Integration of pTau181 with six metabolite features yielded the highest prognostic performance (AUC 0.91; accuracy 80%), with metabolomic findings independently validated in the OPTIMA cohort. A complementary GFAP-NMR panel also performed strongly (AUC 0.80; accuracy 75%). In contrast, individual metabolites, including the atrophy marker homocysteine, and standalone protein biomarkers performed poorly (AUC ≤0.66), as well their combination (AUC 0.68), highlighting the added value of multi-omic integration. In an asymptomatic ageing population (UK Biobank), the models served as a population-level stress test, confirming that multi-omic integration improved specificity for MRI-derived atrophy measures and captured atrophy-related risk in community cohorts.
Conclusion
Multi-omic integration of protein and metabolic features markedly improved prognostication of MCI progression by capturing early neurodegenerative signatures, yielding translational panels suitable for scalable risk stratification and early therapeutic intervention in clinical practice.
