Hypothesis

Integrating proteogenomic and metabolomic data uncovers a causal mitochondrial protein-metabolite axis that drives early neuronal loss in Alzheimer's disease, independent of amyloid burden.

Mechanistic Basis

Recent multi-omics work shows that combined gene-protein expression models outperform polygenic scores in predicting AD risk [3] and that proteogenomic Mendelian randomization identifies circulating proteins with causal roles in arteriosclerosis [4]. We propose that specific mitochondrial enzymes, whose genetic variants alter protein abundance, modulate key metabolites such as acetyl‑CoA and NAD⁺, creating a feed‑forward loop that impairs synaptic energy homeostasis. This axis would be detectable in cerebrospinal fluid (CSF) before measurable cognitive decline and would correlate with proteomic signatures of oxidative phosphorylation complexes.

Testable Predictions

1. Individuals harboring AD‑risk variants in genes encoding mitochondrial complex I subunits will show elevated CSF acetyl‑CoA and reduced NAD⁺ levels compared with non‑carriers, even when amyloid PET is negative.
2. Proteogenomic network models that integrate genotype‑protein‑metabolite edges will predict future cognitive decline better than models using genomics or proteomics alone.
3. Pharmacological restoration of NAD⁺ in variant carriers will normalize the metabolite imbalance and slow cognitive decline in a preclinical model.

Experimental Design

• Cohort: 500 cognitively normal adults aged 55‑80, genotyped for AD‑risk loci, with longitudinal CSF sampling and neuropsychological testing over 3 years.
• Omics: Quantitative proteomics (TMT‑labelled) and targeted metabolomics (acetyl‑CoA, NAD⁺, ATP) on baseline CSF.
• Analysis: Build a Bayesian graph neural network that links genotype → protein abundance → metabolite concentration → cognitive score. Compare model performance (AUC, R²) against polygenic score and proteome‑only predictors.
• Validation: In a subset of 50 participants with elevated risk profile, administer a NAD⁺ precursor (e.g., nicotinamide riboside) for 12 months and assess changes in metabolite levels and cognitive trajectory.

Potential Implications

If validated, this hypothesis would shift early‑intervention strategies from amyloid‑centric approaches to mitochondrial metabolic rescue, providing a biomarker‑driven, mechanism‑based avenue for precision neurology. It also demonstrates how multi‑omics integration can transform correlative findings into actionable, falsifiable targets, addressing the translational gap highlighted in recent reviews [7,8]