Compounding of rare pathogenic copy-number variants and polygenic background is consistent with assortative mating.

This study demonstrates that polygenic scores significantly contribute to the variable expressivity of rare pathogenic copy-number variants and reveals that pervasive assortative mating, rather than collider bias, drives a positive correlation between these rare variants and polygenic background, leading to compounded genetic effects across mutational classes.

The Gist

The Big Picture: Why Do Some People Get Sick and Others Don't?

Imagine you have a genetic lottery ticket. Most people have a ticket with a few "bad numbers" (common genetic variants) that slightly increase their risk of certain health issues, like being a bit heavier or having slightly higher blood pressure. This is your Polygenic Score (PGS)—it's like the background noise of your genetic risk.

Then, imagine someone gets dealt a huge, glaring red card on their ticket. This is a Copy-Number Variant (CNV). It's a large chunk of DNA that is either missing (deletion) or duplicated. These are rare, powerful mutations. Usually, if you get this red card, you expect to have a specific health problem, like a developmental delay or a heart issue.

The Mystery:
But here is the puzzle: Not everyone with the "red card" gets sick. Some have severe symptoms, some have mild ones, and some have almost no symptoms at all. Why?

This paper asks: Does the "background noise" (your PGS) change how loud the "red card" (the CNV) sounds?

The Main Findings

1. The "Volume Knob" Effect (Additive Impact)

The researchers found that your background genetic risk acts like a volume knob for the red card.

• The Analogy: Imagine the red card is a siren. If you have a "protective" background (low risk PGS), it turns the siren down. You might still hear it, but it's manageable. If you have a "risky" background (high risk PGS), it turns the siren up to maximum volume.
• The Result: They looked at 119 different red cards and 43 different health traits. In about 38% of cases, the background genetics made the red card's effect much stronger.
  • Example: A specific missing DNA chunk usually makes people shorter. But if that person also has a genetic background that naturally makes them shorter, they end up being extremely short. If they have a background that makes them tall, they might just be average height.

2. The "Double Trouble" Surprise (Interactions)

Sometimes, the background noise doesn't just turn the volume up; it changes the type of sound.

• The Analogy: It's like mixing two chemicals. Individually, they are mild. But when mixed, they create an explosion.
• The Result: They found two specific cases (involving liver enzymes and grip strength) where having the red card and the risky background created a much worse outcome than just adding the two risks together. This is rare, but it happens.

3. The "Dating App" Mystery (Assortative Mating)

This is the most fascinating part. The researchers noticed something strange: People who carry a "bad" red card often also carry a "bad" background score. They shouldn't necessarily go together.

• The Hypothesis: Why would a person with a rare, bad mutation also have a lot of common bad genes?
• The Explanation: Assortative Mating. This is a fancy term for "birds of a feather flock together."
  • The Analogy: Imagine a dating app.
    • Person A has a rare genetic mutation that makes them slightly shorter.
    • Person B has a common genetic background that also makes them slightly shorter.
    • They meet, fall in love, and have a baby.
    • The baby inherits both the rare mutation from Mom and the common "short" genes from Dad.
  • The Result: Over generations, these two types of "bad luck" get bundled together. The paper shows that this "dating pattern" explains why people with rare mutations often also have high-risk backgrounds. It's not a coincidence; it's because their parents were likely similar in traits, leading to a "double dose" of risk in the child.

Why Does This Matter?

1. Better Predictions: Doctors can't just look at the "red card" (the rare mutation) to predict how sick a patient will be. They need to check the "volume knob" (the polygenic score) too. This helps identify who is at the highest risk and needs the most help.
2. Understanding the "Why": It explains why some families have very sick children while others with the same mutation have healthy kids. It's the combination of the rare mutation and the family's common genetic background.
3. The "Healthy" Bias: The study also looked at why people join big health studies (like the UK Biobank). Usually, sick people don't join. The researchers expected that people with bad mutations who did join would have "super-protective" backgrounds to compensate. Instead, they found the opposite: the "bad" and "bad" were sticking together, proving that the "dating app" (assortative mating) effect is very strong.

In a Nutshell

Think of your health as a house.

• The CNV is a cracked foundation (a rare, big problem).
• The PGS is the quality of the bricks and mortar (common, small problems).

This paper tells us that if you have a cracked foundation, the quality of your bricks matters a lot. If your bricks are also weak, the house might collapse. If your bricks are strong, the house might stand.

Furthermore, the reason many houses have both a cracked foundation and weak bricks is that the parents who built them were likely similar to each other, passing down both problems together. Understanding this helps us build better models for predicting disease and helping people stay healthy.

Technical Summary

1. Problem Statement

Copy-number variants (CNVs) are large genomic deletions or duplications known to cause severe developmental and psychiatric disorders. However, carriers of the same pathogenic CNV often exhibit variable expressivity (different severity) and incomplete penetrance (some remain unaffected). While the "two-hit" model suggests rare secondary variants contribute to this variability, the role of the polygenic background (the cumulative effect of common single-nucleotide variants, or SNVs) remains less understood.

Furthermore, there is a theoretical debate regarding the correlation between an individual's CNV carrier status and their Polygenic Score (PGS):

• Hypothesis A (Participation Bias): In healthy volunteer biobanks (like UK Biobank), deleterious CNV carriers who participate might have a "protective" PGS that counteracts the CNV, leading to a negative correlation.
• Hypothesis B (Linkage Disequilibrium): The PGS might tag the CNV due to local linkage, leading to a positive correlation.
• Hypothesis C (Assortative Mating): Phenotypic similarity between parents (one with a high PGS, one with a CNV) could lead to the co-inheritance of both risk factors in offspring, creating a positive correlation.

This study aims to quantify the additive and interactive effects of CNVs and PGS on a broad spectrum of traits and determine the mechanism driving the observed correlation between CNV status and PGS.

2. Methodology

The study utilized data from the UK Biobank (UKBB), focusing on White British participants.

• Data Cohorts:
  • Training Set ($n = 264,372$): Used to derive genome-wide Polygenic Score (PGS) weights for 43 quantitative phenotypes.
  • Test Set ($n = 96,716$): A CNV-carrier-enriched set of unrelated individuals used to assess the joint impact of CNVs and PGS.
• CNV Data: Analyzed 119 genome-wide significant CNV-trait associations involving 27 CNV regions and 43 phenotypes (metabolic, musculoskeletal, neuropsychiatric, etc.). CNVs were called using PennCNV from genotyping microarrays.
• Statistical Modeling:
  • Additive Effects: Linear regression models were fitted to assess the independent and joint contributions of CNV status and PGS to phenotypic variance.
  • Interaction Effects: Interaction terms between CNV status and PGS were tested on both adjusted (inverse-normal transformed) and raw phenotypic scales.
  • Correlation Analysis: The study calculated the correlation between CNV carrier status and PGS ($r_{CNV, PGS}$).
  • Mechanism Dissection:
    • Cis-Tagging: PGS was split into PGS_cis (SNVs within $\pm$ 50–250 kb of the CNV) and PGS_trans (all other SNVs) to test if local linkage explained the correlation.
    • Assortative Mating: The authors modeled the indirect path of correlation as $r_{indirect} = \alpha \times \beta \times \gamma$, where:
      • $\alpha$: Correlation between CNV and phenotype.
      • $\beta$: Within-couple phenotypic correlation (assortative mating strength).
      • $\gamma$: Correlation between PGS and phenotype.
  • Inheritance Rate: Estimated CNV inheritance rates using Identity-by-Descent (IBD) analysis in related individuals to validate the feasibility of assortative mating.
  • Sensitivity Analyses: Performed permutation tests, adjusted for covariates (sex, age, PCs, residence), and tested robustness against "Winner's curse."

3. Key Contributions

1. Quantification of Additive Effects: Demonstrated that polygenic background significantly contributes to the variable expressivity of rare CNVs across a wide range of quantitative traits.
2. Discovery of Scale-Dependent Interactions: Identified specific cases where the interaction between CNV and PGS is non-additive (epistatic), but only detectable on raw phenotypic scales, not adjusted scales.
3. Mechanistic Insight into Correlation: Provided strong evidence that assortative mating, rather than participation bias or simple linkage disequilibrium, is the primary driver of the positive correlation between rare CNV carrier status and polygenic risk.
4. Generalization: Extended these findings to genome-wide CNV burdens and rare loss-of-function (pLoF) variants, suggesting a pervasive mechanism of genetic compounding.

4. Key Results

• Additive Contribution:

  • For 45 out of 119 (38%) CNV-trait pairs, the PGS had a significant additive effect on the phenotype among CNV carriers.
  • Example: In 16p11.2 BP4-BP5 deletion carriers, the PGS for height significantly exacerbated the height reduction caused by the CNV. Similarly, for 22q11.23 duplications, a high PGS for Gamma-Glutamyltransferase (GGT) led to abnormally elevated GGT levels in 71% of carriers in the top PGS quintile.
  • Clinical Relevance: Integrating PGS with CNV status helps identify carriers at highest risk for severe comorbidities.

• Interactions:

  • Only two significant interactions were found (both on the raw scale): 22q11.23 duplication with GGT levels and grip strength. In these cases, the polygenic background exacerbated the CNV effect more than expected under a purely additive model.
  • Power analysis suggested that most lack of detected interaction is due to limited sample size for rare variants.

• CNV-PGS Correlation:

  • Contrary to the "healthy volunteer bias" hypothesis (which predicts a negative correlation), the study observed a widespread positive correlation between CNV carrier status and PGS for trait-increasing CNVs (76% of pairs were directionally concordant).
  • Cis-Tagging: Local linkage disequilibrium (tagging) explained the correlation for only one common CNV (RHD deletion). For the majority, PGS_trans remained correlated with CNV status, ruling out local tagging as the primary cause.

• Assortative Mating Mechanism:

  • The observed positive correlation ($r_{total}$) strongly matched the correlation predicted by the assortative mating model ($r_{indirect} = \alpha \times \beta \times \gamma$).
  • Result: $r = 0.45$ ($p = 2.0 \times 10^{-7}$).
  • Inheritance: All 17 testable CNVs showed non-zero inheritance rates (pooled rate: 23% for deletions, 79% for duplications), confirming that these variants can be passed down, allowing assortative mating to accumulate risk factors in offspring.

• Broader Mutational Classes:

  • Similar positive correlations were observed between CNV burden and PGS, and between pLoF burden and PGS, suggesting this compounding mechanism is a general feature of the human genome.

5. Significance

• Refining Risk Prediction: The study establishes that PGS is a critical modifier of CNV expressivity. Clinically, this means that a patient's polygenic background should be considered when interpreting the severity of a pathogenic CNV, improving risk stratification for comorbidities.
• Evolutionary and Genetic Architecture: The findings challenge the view that rare and common variants act independently. Instead, assortative mating acts as a mechanism that "compounds" genetic risk, potentially leading to genetic anticipation (increasing disease liability over generations) and explaining why some families show severe phenotypes while others with the same CNV do not.
• Methodological Impact: The study highlights the importance of analyzing raw phenotypic scales to detect interactions and demonstrates that population biobanks are robust against selection bias for detecting these specific genetic correlations, unlike case-control studies which may suffer from collider bias.
• Future Directions: The authors call for larger, diverse biobanks and whole-genome sequencing to better characterize structural variants and refine inheritance rate estimates, which are crucial for modeling the long-term impact of assortative mating on population health.

In conclusion, this paper provides a comprehensive framework showing that the phenotypic outcome of rare pathogenic CNVs is not determined in isolation but is significantly shaped by the polygenic background, a relationship largely driven by the evolutionary force of assortative mating.