Overview

• Peer‑reviewed research in Science Advances integrated temporal, genetic, proteomic, and clinical data across UK Biobank, SAIL, and FinnGen, covering more than 502,000 clinical records and over 52,000 proteomes.
• Higher future risk was linked to diagnoses such as diabetes, vitamin D deficiency, electrolyte imbalance, functional intestinal disorders, noninfective colitis, gastritis, esophagitis, and dyspepsia, while some conditions like diverticular disease showed lower risk in replication analyses.
• Elevated hazards were detectable up to 10–15 years before Alzheimer’s or Parkinson’s diagnosis, with timing patterns varying by condition, including earlier and persistent signals for type 2 diabetes and functional bowel disorders.
• Proteomic markers including neurofilament light chain, GFAP, and peroxiredoxin 1 improved case classification, and machine‑learning ranked age and blood proteins above diagnosis codes for risk prediction.
• Comorbid cases carried a lower average polygenic risk score with no synergistic interaction between genetics and diagnoses detected, and the authors noted limits from UK‑only proteomics, ICD‑10 code reliance, cohort differences, and predominantly European ancestry.