In a Qatar Biobank study of 778 adults, researchers linked plasma proteomics and explainable machine learning to stage 1 hypertension, identifying 36 significantly associated proteins and a best-performing CatBoost model with an AUROC of 0.7985. The study found lower Renin, sRAGE, ghrelin, and IL-1RAcP and higher TFPI, QORL1, HSP70, and C5a in stage 1 hypertension, pointing to candidate blood biomarkers for earlier cardiovascular risk detection.

Why It Matters To Your Practice

• Stage 1 hypertension often precedes overt Heart disease, but blood-based markers that could sharpen early risk assessment remain limited.

• This work suggests proteomic signatures may eventually complement office blood pressure readings by identifying biologic changes tied to vascular dysfunction and oxidative stress.

• Explainable AI may help translate large-scale lab data into clinically interpretable signals rather than black-box predictions alone.

Clinical Implications

• The identified proteins are not ready for routine care, but they highlight candidate biomarkers that could support earlier detection or phenotyping after external validation.

• Renin, TFPI, sRAGE, QORL1, ghrelin, HSP70, IL-1RAcP, and C5a emerged as the most notable candidates for future assay development.

• If replicated, such panels could help clinicians distinguish patients with early biologic hypertension risk even when conventional measures are borderline.

Insights

• Researchers analyzed 1,305 proteins from 778 participants, including 554 controls and 224 stage 1 hypertension cases, with adjustment for age and sex.

• Among tested classifiers, CatBoost performed best with an AUROC of 0.7985.

• SHapley Additive exPlanations were used to interpret model outputs, linking predictive performance to specific proteins and pathways.

• Pathway and network analyses pointed to oxidative stress and vascular function as key biologic themes.

The Bottom Line

• This biobank study links explainable AI and proteomics to smarter early hypertension detection, but the findings remain hypothesis-generating until validated in independent, diverse cohorts.

• For now, clinicians should view these proteins as promising research signals, not practice-ready biomarkers.