Cross-ancestry evaluation of idiopathic pulmonary fibrosis genetic risk variants

This study evaluates the transferability of 35 known idiopathic pulmonary fibrosis genetic risk variants in non-European populations, finding that while most show limited cross-ancestry consistency, the MUC5B variant remains a dominant signal, underscoring the need for larger, diverse studies to improve genetic discovery and translation.

The Gist

The Big Picture: A Recipe Book with Missing Pages

Imagine that Idiopathic Pulmonary Fibrosis (IPF) is a complex disease where the lungs slowly turn into scar tissue, making it hard to breathe. For a long time, scientists have been trying to find the "recipe" for this disease in our DNA.

Think of our DNA as a massive cookbook. Scientists have already found 35 specific "ingredients" (genetic variants) in this cookbook that seem to increase the risk of getting IPF. However, there's a big problem: they only tested these recipes on people of European descent.

It's like a chef who has perfected a cake recipe using only white flour, sugar, and eggs, and then assumes that recipe will work perfectly if you swap in coconut flour, plantains, or yams without checking. This study asked: "Does this recipe work for everyone, or does it fall apart when we try it with different ingredients?"

The Experiment: Testing the Recipe on New Kitchens

The researchers gathered a group of patients with IPF from different parts of the world (Africa, South Asia, East Asia, and the Americas) and compared them to healthy people from the same backgrounds. They also included a specific group from the Basque Country in Spain (who are European but genetically distinct) to see if even small genetic differences mattered.

They took the 35 "ingredients" (genetic variants) known to cause trouble in Europeans and tried to see if they caused trouble in these other groups.

The Results: What Worked and What Didn't

Here is what they found, broken down into simple concepts:

1. The "Star Ingredient" (MUC5B)
There is one specific genetic variant, called rs35705950 (located in a gene called MUC5B), that is the "super-star" of this disease.

• The Analogy: Imagine a specific spice that makes a dish taste terrible. This study found that no matter where you are from—whether you are from London, Lagos, or Lima—this specific spice is the main reason the dish goes bad. It was the most consistent signal across all groups.

2. The "Missing Ingredients" (Monomorphic Variants)
Some of the 35 ingredients found in Europeans simply didn't exist in other populations.

• The Analogy: It's like a recipe calling for "truffles," but when you go to the grocery store in a different country, they don't sell truffles at all. In the East Asian group, for example, 11 of the 35 "ingredients" were completely missing. This means the European recipe book has "blank pages" for these groups.

3. The "Confusing Switch" (PTPN14 Variant)
One specific genetic switch behaved strangely. In Europeans, having this switch usually lowers the risk of disease. But in the African group, having the exact same switch actually increased the risk!

• The Analogy: Think of a light switch. In one house, flipping it "Up" turns the light on. In another house, flipping the same switch "Up" turns the light off. This suggests that the genetic "wiring" works differently depending on your ancestry.

4. The "Prediction Score" (Polygenic Risk Score)
Scientists often create a "Risk Score" by adding up all the bad ingredients to predict who might get sick.

• The Result: This score worked well for European and American groups. However, for African and South Asian groups, the score was much less accurate.
• The Analogy: It's like using a weather app built for London to predict rain in the Amazon. It might get the general idea right, but the specific details are off because the app was trained on the wrong data. When they removed the "Star Ingredient" (MUC5B) from the score, the prediction failed almost entirely for the African group, showing that the other "ingredients" in the recipe don't translate well yet.

Why This Matters (The "So What?")

The main takeaway is that genetics is not one-size-fits-all.

• The Problem: Most genetic research has been done on white Europeans. If we only use those findings to treat people from other backgrounds, we might miss the real causes of their disease or give them the wrong risk assessments.
• The Solution: We need to write new chapters in the genetic cookbook for African, Asian, and other populations. We need more data from diverse groups to understand the full picture.

The Limitations (The "Fine Print")

The authors are honest about the hurdles:

• Small Sample Sizes: They didn't have enough people in the non-European groups to be 100% sure of every result (like trying to judge a whole movie by watching only 5 minutes of it).
• Data Gaps: Some genetic data was hard to read or missing entirely for certain groups, which made the analysis tricky.

The Bottom Line

This study is a crucial first step. It confirms that while one major genetic factor (MUC5B) is a universal villain for lung scarring, the rest of the genetic "villains" are very specific to your ancestry. To cure and treat IPF fairly for everyone, science needs to stop looking at just one type of cookbook and start reading the recipes from all over the world.

Technical Summary

1. Problem Statement

Idiopathic Pulmonary Fibrosis (IPF) is a fatal lung disease with a known genetic component. Previous Genome-Wide Association Studies (GWAS) have identified 35 common genetic risk loci associated with IPF susceptibility. However, these discoveries are heavily biased toward individuals of European genetic ancestry.

• The Gap: There is a lack of validation of these risk variants in non-European populations.
• The Limitation of Existing Data: A recent multi-ancestry GWAS (Partanen et al., 2022) utilized electronic health records (EHR) for case definition, which can lead to imprecise effect estimates due to misclassification.
• Objective: This study aims to evaluate the effects of the 35 known IPF risk variants in clinically curated, non-European individuals to assess transferability, identify ancestry-specific effects, and determine the portability of Polygenic Risk Scores (PRS).

2. Methodology

The study employed a rigorous cross-ancestry association analysis using clinically confirmed IPF cases and matched controls.

• Cohorts & Data Sources:

  • Cases: Clinically curated IPF cases from diverse cohorts (including registries and trials) and an isolated European population (Basque Country).
  • Controls: Matched controls from the UK Biobank (for non-European groups) and the National DNA Bank of Spain (for Basque cases).
  • Ancestry Groups: African (AFR), Admixed American (AMR), East Asian (EAS), South Asian (SAS), and European (EUR).
  • Sample Sizes: 41 AFR, 61 SAS, 21 EAS, 92 AMR, and 98 EUR cases.

• Genetic Processing:

  • Ancestry Assignment: A Random Forest classifier trained on 1000 Genomes Project super-populations (using the first 10 Principal Components) was used to assign ancestry with >50% probability.
  • Quality Control: Relatedness screening removed one individual from pairs with a kinship coefficient $\ge$ 0.0884.
  • Imputation: Performed using the TOPMed R3 reference panel on the Michigan Imputation Server. Variants with low imputation quality ($r^2 < 0.6$) were excluded.
  • Variant Selection: 35 independent genome-wide significant variants from previous IPF studies were selected.

• Statistical Analysis:

  • Association Testing: Logistic regression (PLINK2) adjusted for the first 10 PCs to test variant-IPF associations within each ancestry group.
  • Heterogeneity: Trans-ethnic meta-regression (MR-MEGA) was used to assess heterogeneity of effect sizes across populations (Bonferroni corrected $p < 1.4 \times 10^{-3}$).
  • Polygenic Risk Score (PRS): Constructed using PRSice-2 to assess cumulative genetic risk. Variance explained ($R^2$) was calculated using Nagelkerke's pseudo-$R^2$. Analyses were run both with and without the MUC5B variant.

3. Key Results

• Variant Transferability & Monomorphism:

  • Significant differences in allele frequencies and imputation quality led to the exclusion of variants in specific populations.
  • EAS Cohort: 11 of 35 variants were excluded (7 due to low imputation, 4 monomorphic). Specifically, rs79684490 (10q25.1) and rs116483731 (SPDL1) were monomorphic in EAS.
  • PSKH1 (rs539683219): Previously reported as EAS-specific, this variant was polymorphic only in EAS in this study but showed no significant association with IPF risk, likely due to small sample size and imprecise estimates.

• Consistency of MUC5B:

  • The promoter variant rs35705950 in the MUC5B gene emerged as the dominant and most consistent signal across all ancestry groups, despite substantial differences in allele frequency. It remains the strongest genetic risk factor for IPF globally.

• Ancestry-Specific Heterogeneity:

  • PTPN14 (rs12096551): This was the only variant showing significant heterogeneity in effect sizes ($P = 6.2 \times 10^{-4}$).
    • In Europeans, the C-allele was associated with decreased risk.
    • In Africans (AFR), the same C-allele was associated with increased risk (OR = 3.10, 95% CI: 1.75–5.49).

• Polygenic Risk Score (PRS) Performance:

  • The PRS was associated with IPF risk in all groups, but the variance explained ($R^2$) varied significantly:
    • EUR: $R^2 = 0.09$ (Strongest)
    • AMR: $R^2 = 0.08$
    • SAS & EAS: $R^2 = 0.05$
    • AFR: $R^2 = 0.03$ (Weakest)
  • Impact of MUC5B Removal: Excluding the MUC5B variant drastically reduced the predictive power of the PRS in EUR, AMR, and SAS, and negated the association entirely in AFR ($P=0.133$). In EAS, the association remained significant even without MUC5B.

4. Significance and Contributions

• Validation of MUC5B: Confirms that MUC5B rs35705950 is a universal driver of IPF risk across diverse ancestries, reinforcing its utility as a primary biomarker.
• Evidence of Ancestry-Specific Effects: The study provides concrete evidence that genetic effects are not always transferable. The reversal of the PTPN14 effect in African populations highlights the danger of applying European-derived effect sizes to other groups without validation.
• Limitations of Current PRS: The study demonstrates that current PRS models, heavily reliant on European data and the MUC5B variant, have poor portability to African and South Asian populations. This suggests that non-European IPF risk is driven by different genetic architectures or that current scores fail to capture ancestry-specific linkage disequilibrium (LD) patterns.
• Call for Diversity: The findings underscore the critical need for larger, clinically curated cohorts in understudied populations to improve the generalizability of genetic discoveries and ensure equitable translation of genomic medicine.

5. Limitations

• Sample Size: Modest sample sizes, particularly in EAS (n=21) and AFR (n=41), limited statistical power and precision.
• Variant Bias: The study was restricted to variants previously discovered in European-centric GWAS, potentially missing novel ancestry-specific risk loci.
• Imputation Quality: Variability in imputation quality across ancestries reduced the number of analyzable variants, particularly in the EAS cohort.
• PRS Construction: The PRS did not account for ancestry-specific effect sizes or LD patterns, which likely contributed to the reduced performance in non-European groups.