Integrative Proteomic and Metabolomic Signatures of Accelerated PhenoAge in the UK Biobank

By integrating plasma proteomics and metabolomics data from approximately 20,000 UK Biobank participants, this study identifies coordinated multi-omics modules—dominated by immune-inflammatory activation with metabolic support—that characterize accelerated PhenoAge and offer enhanced biological interpretability beyond single-omics analyses.

The Gist

Imagine your body is a massive, bustling city. For years, scientists have tried to figure out why some cities (people) stay vibrant and functional for decades, while others start to crumble, suffer from traffic jams, and lose power much earlier than expected.

This paper is like a team of detectives using two different types of high-tech scanners to investigate why some people's "cities" are aging faster than their actual birth date suggests.

Here is the story of their investigation, broken down simply:

1. The Two Scanners: Proteins and Metabolites

The researchers looked at blood samples from about 20,000 people in the UK. They used two different "scanners" to read the city's status:

• The Protein Scanner (Proteomics): This looks at the "workers" and "machines" in the city. Proteins are the actual tools doing the work (building, repairing, fighting infections).
• The Metabolite Scanner (Metabolomics): This looks at the "fuel" and "waste." Metabolites are the chemicals left over from burning fuel, the energy being used, and the trash being produced.

They compared these scans against a "Biological Age Clock" called PhenoAge. This clock doesn't just ask, "How old are you?" It asks, "How well are your organs working?" If your organs act like they are 65 but you are only 50, you have "accelerated aging."

2. The Detective Work: Who is Aging Fast?

The team didn't look at everyone. They picked the "extremes":

• Group A: People whose biological age was much older than their real age (the "fast agers").
• Group B: People whose biological age was younger than their real age (the "slow agers").

They wanted to see what molecular differences existed between these two groups.

3. The Findings: The Protein Scanner Wins (But the Fuel Scanner Helps)

When they looked at the data alone, the Protein Scanner was the clear winner. It could easily tell the difference between the fast agers and slow agers. It was like looking at a city and immediately seeing that the construction crews were exhausted and the police were overwhelmed.

The Metabolite Scanner (the fuel/waste) wasn't as good at telling them apart on its own. However, when they combined the two scanners, they didn't just get a better score; they got a better story.

4. The Four "Neighborhoods" of Aging

By combining the data, the researchers found four distinct "neighborhoods" or patterns in the body that were different in the fast agers. Think of these as four different districts in the city that were in trouble:

• District 1: The Fire Department (Immune & Inflammation)

  • What's happening: The body is in a constant state of low-grade alarm. The "firefighters" (immune cells) are running around shouting, and the "smoke" (inflammation) is everywhere.
  • The Clue: High levels of inflammatory proteins and "smoke" markers (like GlycA). This is the strongest sign of aging. It's like a city where the sirens are always blaring, wearing everyone out.

• District 2: The Highway System (Lipids & Blood Vessels)

  • What's happening: The roads are getting clogged. The way fats and cholesterol move through the blood is getting messy.
  • The Clue: Issues with how the body transports fats (lipids) and how blood vessels signal each other. It's like a city where the delivery trucks are stuck in traffic, causing bottlenecks.

• District 3: The Power Plant (Nutrients & Energy)

  • What's happening: The energy supply is fluctuating. The city is struggling to manage its fuel (sugar, amino acids like glutamine) and minerals (like iron and phosphate).
  • The Clue: Changes in how the body processes nutrients. It's like a power plant that is sputtering, unable to keep the lights on steadily.

• District 4: The Maintenance Crew (Protective Factors)

  • What's happening: This is the "good" neighborhood. In the slow agers, the body has a strong team of "maintenance workers" (specifically HDL cholesterol and certain proteins) that keep things clean and running smoothly.
  • The Clue: These people have better "trash collection" and "road repair" systems, keeping their biological age low.

5. The Big Takeaway

The most important lesson from this paper is this: Looking at just one part of the city isn't enough.

If you only looked at the workers (proteins), you'd know the city is tired. But if you also look at the fuel and waste (metabolites), you understand why they are tired. Maybe the power plant is failing, or the trash isn't being collected, causing the workers to burn out.

In simple terms:
Aging isn't just one thing going wrong. It's a complex chain reaction where inflammation, clogged roads, and bad fuel management all feed into each other. By using both scanners together, scientists can now see the whole picture of how these systems break down, which helps them figure out how to fix them later.

Why does this matter?

This study gives doctors and scientists a new "map" of aging. Instead of just guessing why someone is aging fast, they can now see specific "neighborhoods" that need repair. Maybe in the future, we won't just treat old age; we might be able to target the specific "traffic jam" or "power outage" causing the problem, helping people stay healthy for longer.

Technical Summary

1. Problem Statement

Aging is a heterogeneous process where individuals of the same chronological age exhibit varying physiological functions, disease burdens, and mortality risks. While "aging clocks" (e.g., epigenetic, proteomic, metabolomic) have been developed to quantify biological age, the underlying molecular mechanisms driving inter-individual variation remain poorly understood.

• The Gap: Most existing studies focus on single-omics layers (e.g., proteomics or metabolomics) or prioritize predictive accuracy over biological interpretability.
• The Objective: To identify coordinated molecular processes across proteomic and metabolomic layers that distinguish individuals with accelerated biological aging from those with slower aging, using a large-scale, clinically grounded metric (PhenoAge).

2. Methodology

Study Population & Data Source

• Cohort: A subcohort of the UK Biobank (~20,000 participants).
• Selection Criteria: Participants with complete plasma proteomic (Olink), metabolomic (Nightingale NMR), and clinical data required for PhenoAge calculation.
• Final Sample: 18,632 participants (mean age 56.7 years; 54.1% female).
• Phenotype Definition: Biological aging was measured using PhenoAge, a composite biomarker derived from 9 clinical blood markers and chronological age. PhenoAgeAccel (acceleration) was calculated as the residual of PhenoAge regressed on chronological age.
• Stratification: To enhance biological contrast, the study focused on the extreme 20% tails of the PhenoAgeAccel distribution (Accelerated vs. Not Accelerated groups).

Data Preprocessing

• Covariate Adjustment: Chronological age and sex were regressed out of the omics data prior to modeling.
• Circularity Control: Metabolites directly used in the PhenoAge calculation (creatinine, albumin, glucose) were excluded from the metabolomic block to prevent circularity.
• Imputation: Missing values were imputed using k-nearest neighbors; metabolites were log1p-transformed.

Statistical Framework

• Single-Omics Analysis: Sparse Partial Least Squares Discriminant Analysis (sPLS-DA) was applied separately to proteomic and metabolomic data to assess individual discriminative power.
• Multi-Omics Integration: The DIABLO (Data Integration Analysis for Biomarker discovery using Latent cOmponents) framework (from the mixOmics R package) was used. This supervised multivariate method identifies correlated features across different data types (proteins and metabolites) that jointly discriminate the outcome groups.
• Validation: Models were evaluated using Balanced Error Rates (BER) and Area Under the Curve (AUC) via cross-validation and an independent test set.
• Biological Validation: Selected proteins were cross-referenced against curated aging databases (GenAge, CellAge, LongevityMap, GenDR, SASP Atlas). Functional enrichment and network analysis (correlation circles, Circos plots, relevance networks) were performed.

3. Key Results

Model Performance

• Single-Omics: Proteomics alone provided strong discrimination between aging groups (BER = 0.15). Metabolomics showed weaker performance (BER = 0.233) and required more components for separation.
• Integrative Model: DIABLO did not significantly improve classification accuracy over proteomics alone (Train BER = 0.176; Test BER = 0.168). However, it successfully revealed four distinct, biologically interpretable multi-omics modules (latent components).

The Four Multi-Omics Modules

1. Immune-Inflammatory Axis (Component 1):
   • Dominant Signal: The strongest separation between groups.
   • Features: High loadings of inflammatory proteins (OSM, TNFRSF1A, IL6, GDF15, MMP9) and growth factors (HGF, IGFBP4).
   • Metabolites: Glycoprotein acetyls (GlycA) and Glucose:Lactate ratio.
   • Interpretation: Represents a pro-inflammatory immune-metabolic state consistent with "inflammaging."
2. Lipid-Vascular Axis (Component 2):
   • Features: Polyunsaturated fatty acids (PUFAs), cholesteryl esters in LDL, and proteins involved in receptor tyrosine kinase signaling (ERBB2/3, EGFR, FGFR2), lipid homeostasis (LDLR, TTR, LCAT), and complement/coagulation cascades.
   • Interpretation: Links lipid metabolism, vascular signaling, and coagulation, potentially reflecting cardiometabolic risk and amyloidosis pathways.
3. Nutrient-Metabolic Axis (Component 3):
   • Features: Anchored by Glutamine. Proteins include FGF23 (phosphate), TFRC (iron), EPO (erythropoiesis), and stress signaling (CRH).
   • Interpretation: Suggests a regulatory axis involving nutrient sensing, mineral metabolism, and energy production (TCA cycle via $\alpha$-ketoglutarate).
4. HDL-Apolipoprotein Axis (Component 4):
   • Features: Apolipoproteins (APOM, APOA1, APOC1, APOD) and HDL-related metabolites.
   • Interpretation: Enriched in the "Not Accelerated" group, suggesting a protective, lipid-regulatory signature. However, effect sizes were small, indicating limited robustness.

Database Overlap & Novelty

• Validation: 49% of the 100 DIABLO-selected proteins overlapped with established aging/senescence databases (highest overlap in Components 1 and 2).
• Novel Candidates: The study identified proteins with limited prior aging annotation as key discriminators, including PRRT3, VSIG4, LILRA5, and DRAXIN, suggesting new targets for functional investigation.

4. Significance and Conclusions

• Biological Interpretability over Prediction: The study demonstrates that while multi-omics integration may not always yield superior predictive accuracy compared to single-omics (proteomics dominated here), it is superior for uncovering coordinated molecular mechanisms.
• Mechanistic Insight: The integration revealed that accelerated aging is not just a protein-level phenomenon but involves specific metabolic support systems (e.g., the link between inflammation and the glucose:lactate ratio, or lipid transport and vascular signaling).
• Metabolomic Role: Although metabolomics had weaker standalone performance, it played a critical "anchoring" role in the integrated model, linking protein changes to underlying metabolic states and energy pathways.
• Clinical Relevance: The identification of distinct axes (immune, lipid-vascular, nutrient) provides a structured framework for understanding the heterogeneity of aging and potential targets for interventions aimed at extending healthspan.

Limitations Noted:

• Cross-sectional design prevents causal inference.
• Reliance on plasma measurements may miss tissue-specific aging.
• Need for independent validation of novel candidates (e.g., PRRT3).
• Potential confounding by lifestyle factors (smoking, diet) not fully adjusted for in the initial model.

In summary, this paper provides a robust, integrative map of the molecular landscape of accelerated aging, highlighting that the interplay between immune activation, lipid metabolism, and nutrient regulation defines the biological aging process better than any single layer alone.