Development of a Multi-Trait Polygenic Score for Intrinsic Capacity

This study developed and validated a multi-trait polygenic score for intrinsic capacity using data from over 13,000 Canadian participants, demonstrating that a model integrating genetic predictors across five domains effectively captures the multisystem biological basis of healthy aging and is significantly associated with reduced mortality.

The Gist

Imagine your body is like a high-performance car. For a long time, mechanics (doctors) have checked how well this car runs by looking at the visible parts: the speedometer (how fast you walk), the fuel gauge (your energy levels), the engine noise (your hearing), and the dashboard lights (your mood). This is what scientists call Intrinsic Capacity (IC). It's a way to measure how "healthy" and resilient you are as you age, covering five main areas: thinking, moving, feeling, energy, and senses.

However, there's a problem with just looking at the dashboard. By the time a warning light turns on, the engine might already be damaged. The car's "true" health is determined by the microscopic parts inside the engine that we can't see yet.

This paper is about building a genetic "dashboard" that predicts how your car will run before the warning lights even flicker.

The Big Idea: The Genetic Recipe Book

The researchers wanted to create a score based on your DNA (your genetic recipe book) that predicts your overall Intrinsic Capacity.

In the past, scientists tried to make a single score based on just one trait, like "how strong your grip is." But the human body is too complex for just one number. It's like trying to predict how a cake will taste by only measuring the amount of sugar. You need to know about the flour, the eggs, the baking powder, and the oven temperature too.

So, this team decided to mix 63 different genetic ingredients (polygenic scores) related to things like body weight, blood pressure, memory, and mood. They used a smart computer program (Machine Learning) to figure out which of these 63 ingredients actually mattered most for the final "cake" (your overall health).

How They Did It (The Kitchen Test)

1. The Ingredients: They gathered genetic data on 63 different traits from a massive library of scientific studies.
2. The Chef (The Computer): They fed these 63 genetic "ingredients" into a computer chef. The chef's job was to taste-test different combinations to see which mix best predicted the health of 13,000 real people from the Canadian Longitudinal Study on Aging.
3. The Result: The computer found that only 33 of those 63 ingredients were actually needed to make the perfect prediction. It threw out the rest.

What Did They Find?

The computer's "recipe" for a healthy, resilient body had some surprising highlights:

• The "Vitality" Stars: The strongest ingredients in the recipe were things related to energy and physical strength. Think of Body Mass Index (BMI), grip strength, and how much muscle you have. It turns out, your body's "engine" (metabolism and physical power) is the foundation of your overall health.
• The Full Team: But it wasn't just physical strength. The recipe also included ingredients for thinking (like fluid intelligence), mood (like ADHD genetics), and senses (like hearing). This proves that your brain, your heart, and your muscles are all connected in a complex web.
• The Proof: When they tested this new genetic score against the real-world "dashboard" (how people actually felt and moved), the two matched up.
  • The Magic Stat: People with a higher genetic score for "Intrinsic Capacity" were 96% less likely to die in the study period compared to those with a low score. That is a massive difference!

Why Does This Matter?

Think of this new score as a weather forecast for your body.

• Old Way: You wait until you get sick or frail (the storm hits) to realize you need an umbrella.
• New Way: This genetic score tells you, "Hey, your body's engine is genetically prone to rusting in the engine block area (vitality) or the brakes (mobility)."

Because this score is based on your DNA, it doesn't change with your diet or exercise yet. It tells you your starting potential. If you know your genetic "engine" is strong, you can push it harder. If you know it's a bit fragile, you can start taking extra care (better diet, more exercise, regular check-ups) years before you actually feel weak.

The Catch

There are a few things to keep in mind:

• The Sample: The study mostly looked at people of European ancestry. Just like a recipe might taste different with different ingredients, this genetic score might need to be tweaked for people from other backgrounds.
• It's Not Magic: The score explains about 2% of the differences in health. That sounds small, but in the world of genetics, it's a huge step forward. It's like finding a new, powerful spice in your kitchen; it doesn't cook the meal for you, but it makes the flavor much better.

The Bottom Line

This study is the first to successfully mix genetics with the concept of "Intrinsic Capacity." It shows that our ability to age well isn't just about one thing; it's a complex mix of our energy, our brain, and our senses. By understanding our genetic blueprint for these traits, we might one day be able to predict and prevent frailty before it ever starts, keeping our "cars" running smoothly for much longer.

Technical Summary

1. Problem Statement

Intrinsic Capacity (IC) is a World Health Organization (WHO) framework defining healthy aging as the composite of an individual's physical and mental capacities across five domains: cognitive, locomotor, psychological, vitality, and sensory. While IC is a robust predictor of frailty, disability, and mortality, current assessment models rely entirely on phenotypic data (clinical assessments and self-reports).

The study addresses two critical gaps:

1. Biological Blind Spots: Phenotypic tools capture observable functional decline but miss the subtle physiological and genetic processes that precede visible changes.
2. Multidimensional Complexity: Traditional single-trait Polygenic Scores (PGS) focus on individual phenotypes, failing to capture the complex, multidimensional biological architecture of IC. Furthermore, it is unclear if one domain (e.g., vitality) underpins the others or if they contribute equally.

The objective was to develop and validate a Multi-Trait Polygenic Score (mt-PGS) that integrates genetic predictors across all five IC domains to create a genetically informed index of intrinsic capacity.

2. Methodology

The study utilized data from the Canadian Longitudinal Study on Aging (CLSA), specifically the comprehensive cohort.

• Study Cohort: 13,085 participants (mean age 61 years, 97.4% European ancestry) with both genetic and phenotypic data.
• Target Outcome: A previously established phenotype-based IC Z-score derived from 14 baseline variables via factor analysis.
• Input Features: 63 distinct phenotypes associated with the five IC domains were selected based on literature review. Genome-Wide Association Study (GWAS) summary statistics for these phenotypes were sourced from public databases (excluding CLSA to prevent sample overlap).
• Polygenic Score Construction:
  • Method: Polygenic Risk Score-Continuous Shrinkage (PRS-CS), a Bayesian regression framework.
  • Process: Used GWAS summary statistics and the 1000 Genomes European reference panel (LD structure) to infer posterior effect sizes.
  • Output: 63 standardized (z-distributed) PGSs were computed for each participant.
• Model Development (Machine Learning):
  • Algorithm: Elastic Net (penalized regression) implemented in R (glmnet).
  • Validation Strategy: Five-fold nested cross-validation to tune hyperparameters ($\alpha$ for mixing L1/L2 penalties and $\lambda$ for regularization strength) and prevent overfitting.
  • Selection: The model selected the optimal subset of PGSs (features) with non-zero coefficients to predict the phenotype-based IC score.
• Validation:
  • Convergent Validity: Correlation between the new mt-PGS IC score and the existing phenotype-based IC score.
  • Predictive Validity: Association with mortality using binary logistic regression (adjusted for age, sex, and top 5 principal components).

3. Key Contributions

• First mt-PGS for IC: This is the first study to construct a multi-trait polygenic score specifically for the holistic concept of Intrinsic Capacity.
• Domain Integration: Successfully integrated genetic predictors from all five WHO-defined IC domains, moving beyond single-trait approaches.
• Variance Explanation: Demonstrated that aggregating diverse polygenic signals significantly improves the prediction of IC compared to single-trait models.
• Biological Insight: Provided empirical evidence regarding the relative contribution of different biological domains to the overall IC construct.

4. Key Results

• Model Performance:

  • Out of 63 input PGSs, 33 phenotypes were retained with non-zero coefficients in the final elastic net model.
  • The mt-PGS model explained 2.23% of the variance in the phenotype-based IC score.
  • This represents a >7-fold increase in variance explained compared to a previous single-trait IC-PGS (which explained ~0.3%).
  • Performance metrics: RMSE = 0.98, MAE = 0.79.

• Top Predictors (Feature Importance):

  • The strongest contributors were predominantly aligned with the Vitality domain, though cross-domain overlap was acknowledged.
  • Top 5 Contributors:
    1. Body Mass Index (BMI) ($\beta = -0.04$)
    2. Grip Strength ($\beta = 0.03$)
    3. Diastolic Blood Pressure ($\beta = 0.017$)
    4. Fat-Free Mass ($\beta = 0.015$)
    5. Chronic Obstructive Pulmonary Disease (COPD) ($\beta = -0.014$)
  • Other Domains: The model also retained strong predictors from Psychological (ADHD), Cognitive (Fluid Intelligence, Processing Accuracy), Sensory (Hearing Impairment), and Locomotion (Difficulty Walking) domains.

• Validity and Mortality Association:

  • Correlation: A statistically significant positive correlation ($r = 0.16$) was found between the mt-PGS IC score and the phenotype-based IC score.
  • Mortality Risk: Higher mt-PGS IC scores were strongly inversely associated with mortality.
    • Odds Ratio (OR): 0.04 (95% CI: 0.005 – 0.379).
    • Interpretation: A one standard deviation (Z-score) increase in the mt-PGS IC score corresponds to an approximate 96% reduction in the odds of death.

5. Significance and Implications

• Biological Validity: The findings support the hypothesis that IC is a multisystem biological trait. The prominence of vitality-related phenotypes (metabolic, cardiovascular, respiratory) suggests that physiological reserve is a foundational component of IC, while still acknowledging the necessity of cognitive and sensory inputs.
• Clinical Potential: While the variance explained (2.23%) is modest in isolation, it is comparable to other complex aging traits. Integrating mt-PGS with phenotypic data could enable earlier identification of individuals at risk for functional decline before overt clinical symptoms appear, facilitating targeted preventive interventions.
• Future Directions: The study establishes a foundation for integrating genetic susceptibility into the WHO IC framework. Future work requires validation in diverse ancestral populations (currently limited to European ancestry) and exploration of how mt-PGS interacts with lifestyle and environmental factors to improve risk stratification.

Limitations: The study is limited by the predominantly European ancestry of the cohort, potential heterogeneity in GWAS sample sizes for the input phenotypes, and the need for external validation in independent cohorts.