Explaining the unexplained admixture mapping signals via rare variants: the HCHS/SOL

Using whole-genome sequencing and metabolomics data from the HCHS/SOL study, this research demonstrates that while rare variants show replicated associations with metabolite levels, expanding the genomic search region for common variants, rather than including rare variants, is the primary factor explaining previously unexplained admixture mapping signals.

The Gist

The Big Picture: Solving a Genetic Mystery in a "Melting Pot"

Imagine the genetic makeup of a person from a Hispanic or Latino background as a giant, delicious stew. This stew is made by mixing ingredients from three main "broths": European, African, and Indigenous American (Amerindian). Because these groups were separated for thousands of years before mixing, they brought different "flavors" (genetic variants) to the pot.

Scientists use a technique called Admixture Mapping to taste this stew and figure out which specific "broth" is responsible for a specific flavor (like a high level of a certain chemical in the blood, known as a metabolite).

The Problem:
In previous studies, scientists found that certain parts of the stew definitely came from a specific broth and caused a specific flavor. However, when they tried to find the exact ingredient (the specific gene variant) causing that flavor, they hit a wall. They looked for the "common ingredients" (common genetic variants) and found some, but they couldn't explain the whole flavor. It was like finding a few carrots in the soup but still not knowing why the soup tasted so spicy.

The New Hypothesis:
The researchers asked: "Maybe the missing flavor isn't caused by common carrots, but by rare, exotic spices that are hard to find?" These are "rare variants"—genetic changes that are very uncommon in the general population but might be concentrated in specific ancestral groups.

The Experiment: A Two-Pronged Search

The team used data from the HCHS/SOL study (a massive health study of Hispanic/Latino people in the US) to test two ideas:

1. The "Rare Spice" Theory: Maybe the unexplained flavors are caused by rare genetic variants that we haven't looked at closely enough.
2. The "Wider Net" Theory: Maybe we just didn't look at a wide enough area of the genome to find the common ingredients that are actually responsible.

They used a high-tech "microscope" (Whole Genome Sequencing) to look at the DNA of over 5,000 people and measured their blood chemistry (metabolites).

The Results: What Did They Find?

1. The Rare Spices Were Real, But Small

They did find some "rare spices." They identified sets of rare genetic variants that were indeed linked to specific blood chemicals.

• The Catch: These rare variants were like finding a single, tiny pinch of saffron. While they existed and were interesting, they only explained a tiny fraction of the mystery. They couldn't account for the massive flavor changes the scientists were seeing.

2. The "Wider Net" Solved the Mystery

When the researchers stopped looking for just the "main" common ingredients and instead cast a much wider net (looking at a larger area of the genome around the signal), they found the answer.

• The Discovery: It turned out that the "unexplained" signals were actually caused by common variants that were just spread out over a larger area than previously thought.
• The Analogy: Imagine you are trying to find the source of a smell in a room. You first look at the center of the room and find nothing. You then expand your search to the whole room, and suddenly you see a pot of soup boiling on the stove. The smell wasn't a mystery; you just hadn't looked at the whole kitchen.

The Conclusion: Common Variants Rule (Mostly)

The study concludes that for the vast majority of these "unexplained" genetic signals in admixed populations, common variants (the standard ingredients) are the real culprits. They just needed to be looked at in a wider context.

While rare variants (the exotic spices) do play a role and are important for understanding specific diseases, they are not the main reason why these specific ancestry-linked signals were previously a mystery.

Why Does This Matter?

• For Science: It tells researchers that when they see a genetic signal they can't explain, they shouldn't immediately jump to "it must be a rare, hidden variant." They should first check if they are looking at a wide enough area to find the common variants.
• For Health: By understanding that these signals are often driven by common variants, we can better predict disease risks and develop treatments that work for diverse populations, not just those of European descent.
• For the "Stew": It helps us appreciate that the "flavor" of our health is often a mix of standard ingredients, not just the rare, exotic ones.

In short: The scientists went looking for a needle in a haystack (rare variants) to explain a mystery, but they realized the mystery was actually caused by a whole pile of hay (common variants) that they just hadn't looked at closely enough.

Technical Summary

1. Problem Statement

Admixture Mapping (AM) is a powerful method for identifying genetic loci associated with traits in admixed populations (e.g., Hispanic/Latino groups) by leveraging differences in local ancestry. However, a significant challenge remains: many statistically significant AM signals cannot be fully explained by adjusting for nearby Genome-Wide Association Study (GWAS) identified common variants.

Previous studies suggested that these "unexplained" signals might be driven by rare variants (RVs) (Minor Allele Frequency < 0.01) that are ancestry-enriched. While common variants often tag causal loci via Linkage Disequilibrium (LD), the extensive length of local ancestry segments in admixed populations makes pinpointing the specific causal variant difficult. The authors aimed to test the hypothesis that rare variants are responsible for AM signals that persist after conditioning on known common variants.

2. Methodology

The study utilized data from the Hispanic Community Health Study/Study of Latinos (HCHS/SOL), a large, diverse cohort with whole-genome sequencing (WGS) and metabolomics data.

• Data Source:

  • Cohort: 5,307 individuals with WGS, local ancestry inference, and metabolomics data.
  • Metabolomics: Two batches (Discovery: $N \approx 3,842$; Replication: $N \approx 1,465$) measuring fasting serum metabolites.
  • Local Ancestry: Inferred using FLARE (Fast Local Ancestry Inference Estimation) based on the TOPMed reference panel, distinguishing African, Amerindian, and European ancestries.

• Study Design:

  • Regions Selected: 16 genomic regions previously identified with significant AM signals for specific metabolites.
    • 4 Negative Controls: Regions where AM signals were already known to be explained by common variants.
    • 12 Test Regions: Regions where AM signals were not explained by common variants in prior studies.
  • Rare Variant Analysis: Adapted the STAAR pipeline (Sequence Kernel Association Test for Rare variants) to perform gene-centric burden tests on coding and non-coding functional categories (e.g., pLoF, missense, enhancers).
  • Conditional Analysis Strategy:
    1. Identified "associated common variants" by combining GWAS catalog hits with new single-variant associations found in the expanded genomic region (±1 Mb).
    2. Extracted burden scores from significant rare variant sets.
    3. Performed Conditional Admixture Mapping using four models:
       • Null model (no variants).
       • Adjusted for common variants only.
       • Adjusted for rare variant burden scores only.
       • Adjusted for both common and rare variants.
  • Goal: Determine if including rare variants (or expanding the common variant search space) caused the AM signal to become non-significant, thereby "explaining" the signal.

3. Key Contributions

• Systematic Evaluation of Rare Variants in AM: This is one of the first comprehensive studies to specifically test whether rare variants explain AM signals that common variants fail to capture in an admixed population.
• Methodological Adaptation: Modified the STAAR pipeline to extract burden scores specifically for use as covariates in conditional admixture mapping, allowing for the direct assessment of rare variant contributions to ancestry-based associations.
• Refutation of the "Rare Variant Hypothesis" for Unexplained Signals: The study provides strong evidence that the "unexplained" AM signals are likely not driven by rare variants, but rather by common variants located further away than previously considered.

4. Key Results

• Rare Variant Discovery:

  • The study identified 35 significant rare variant sets (coding and non-coding) across the 16 regions in the discovery batch.
  • Only 4 of these rare variant sets replicated in the independent batch 2.
  • Crucially, when conditioning on the identified common variants, the number of significant rare variant sets dropped drastically (from 104 to 35 in batch 1), suggesting that common variants account for the majority of the signal.

• Explaining the AM Signals:

  • Negative Controls: As expected, adjusting for common variants eliminated the AM signals.
  • Test Regions (The "Unexplained" Signals):
    • Adjusting for common variants (identified within an expanded genomic window) successfully explained the majority of the AM signals in all 12 test regions. The signals became non-significant or substantially reduced.
    • Adjusting for rare variant burden scores alone had minimal impact on the AM signals. In most regions, the AM signal remained significant after adjusting for rare variants.
    • Combined Model: Adding rare variants to the model that already included common variants did not further reduce the AM signal, indicating that rare variants did not provide additional explanatory power.

• Spatial Distribution:

  • The "associated common variants" that explained the signals often spanned a wide genomic region (1–2 Mb), sometimes extending far beyond the initial AM peak.
  • Significant rare variant sets, when found, were located very close to these common variants, suggesting they tag the same underlying causal gene/locus rather than acting as independent drivers.

• Specific Examples:

  • Propyl 4-hydroxybenzoate sulfate (Chr 16): Rare variants showed some signal reduction, but common variants were the primary driver.
  • 3-aminoisobutyrate (Chr 5) & Ethylmalonate (Chr 12): Rare variants (e.g., in AGXT2 and ACADS) were identified but did not explain the AM signal once common variants were accounted for.

5. Significance and Conclusion

• Primary Conclusion: The "unexplained" admixture mapping signals in the HCHS/SOL cohort are not primarily driven by rare variants. Instead, they are explained by common variants that were missed in previous analyses because they were located outside the narrow genomic windows initially tested.
• Implication for Genetic Studies:
  • Researchers should expand the genomic search window when investigating AM signals before concluding that rare variants are the cause.
  • While rare variants are biologically important, their specific role in explaining ancestry-enriched AM signals appears limited compared to the contribution of common variants in LD with the local ancestry blocks.
  • Admixture mapping remains a valuable tool for identifying ancestry-specific risk loci, but the causal variants are often common variants with ancestry-specific frequency differences rather than rare, ancestry-enriched mutations.
• Future Directions: The authors suggest that future work should focus on fine-mapping these expanded regions and investigating the functional relevance of the identified common and rare variants, acknowledging that sample size limitations may have reduced power to detect very rare variants.