Hypothesis

We hypothesize that a longitudinal multi‑omic network signature—combining baseline plasma proteomics, metabolomics, and transcriptomic shifts measured every three months—predicts the onset of frailty within 12 months in adults aged ≥65 with two or more chronic conditions.

Rationale

Multi‑omics platforms have shown promise in oncology and cardiovascular disease by linking molecular layers to clinical outcomes (1, 2). However, most applications remain static and disease‑specific, overlooking the systemic crosstalk that drives multimorbidity in aging (5). Recent work shows that phosphoproteomic rewiring and metabolite flux changes precede clinical frailty, suggesting a mechanistic bridge between molecular dysregulation and physiological decline (3). We propose that integrating these layers into a dynamic Bayesian network captures emergent system‑level risk that static models miss.

Novel Mechanistic Insight

We suggest that chronic low‑grade inflammation drives coordinated alterations in kinase‑substrate networks (detectable via phosphoproteomics) and perturbs tryptophan‑kynurenine metabolism (detectable via metabolomics). These coupled changes reduce mitochondrial resilience and muscle protein synthesis, accelerating frailty. By modeling the directionality of these interactions—e.g., inflammation → kinase activation → metabolite shift → muscle catabolism—we generate a testable causal chain.

Testable Predictions

• Prediction 1: Individuals whose baseline multi‑omic network score exceeds a defined threshold will have a ≥2‑fold higher hazard of developing frailty (Fried phenotype) within 12 months compared with those below the threshold (hazard ratio >2.0, p<0.05).
• Prediction 2: Targeted intervention (e.g., IL‑6 blockade combined with leucine supplementation) that normalizes the predicted phosphoproteomic‑metabolomic axis will reduce frailty incidence by at least 30% relative to control in a randomized pilot.
• Prediction 3: The network score will outperform single‑omics or clinical‑only models in predicting frailty (increase in AUC ≥0.07) when assessed via cross‑validation.

Falsifiability

If longitudinal sampling shows no significant association between the multi‑omic network score and frailty onset, or if intervening on the predicted axis fails to alter frailty rates, the hypothesis is falsified. Similarly, if adding further omics layers does not improve prediction beyond the proposed network, the claimed mechanistic integration is not supported.

Implementation Sketch

1. Recruit 500 adults ≥65 with ≥2 chronic conditions from a community cohort.
2. Collect plasma, PBMCs, and clinical frailty measures at baseline, 3, 6, 9, and 12 months.
3. Perform phosphoproteomics (LC‑MS/MS), untargeted metabolomics, and transcriptomics (RNA‑seq) on each visit.
4. Construct a dynamic Bayesian network using time‑lagged correlations; derive a risk score.
5. Validate predictions against Fried frailty criteria; test intervention in a subset.