Using Multiplexed Functional Data to Reduce Variant Classification Inequities in Underrepresented Populations

Medicine Genetics Life Sciences
multiplexed functional data variant classification inequity underrepresented populations genome medicine variant effects clinical significance

Title: “Using Multiplexed Functional Data to Reduce Variant Classification Inequities in Underrepresented Populations”

Authors and Publishing Information:

The paper, authored by Moez Dawood et al., is published in the journal Genome Medicine (2024, Volume 16, Issue 143). The authors represent multiple institutions such as Baylor College of Medicine, the University of Washington, and the St Vincent’s Institute of Medical Research. This open-access article complies with the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Overview

Genomic medicine faces substantial challenges regarding health equity, particularly when addressing genetic variations in individuals of non-European ancestry. Due to a longstanding focus on European-ancestry populations in research and clinical sequencing, critical gaps in the understanding of disease risk associated with genetic variants in other populations persist. These disparities often result in higher rates of inconclusive diagnostic outcomes—classified as “Variants of Uncertain Significance” (VUS)—for individuals of non-European ancestry, whereas European-ancestry populations are more frequently categorized with “Pathogenic or Likely Pathogenic” (P/LP) variants. This observed gap introduces clinical complications, including potential misdiagnosis or suboptimal medical management.

To address this inequity, the study proposes the application of Multiplexed Assays of Variant Effects (MAVE) as a solution. MAVE systematically tests all potential single-nucleotide variants (SNVs) and insertions/deletions (indels) in specific genes and generates saturation-style functional data that improves variant reclassification. The study demonstrates that applying MAVE data can significantly reduce VUS disparities, particularly benefiting underrepresented populations.

Research Methodology

1. Data Sources and Sample Groups

The study analyzed genomic data from: - 245,394 individuals enrolled in the All of Us (v7) cohort. - Independent datasets from the Genome Aggregation Database (gnomAD): - 123,709 exomes from gnomAD v2.1.1 - 51,535 genomes from gnomAD v3.1.2, excluding overlapping individuals from v2.

The total sample included 420,638 individuals: - Individuals of European-like ancestry: 213,663 - Individuals of non-European-like ancestry: 206,975

Populations were grouped into “European-like” and “non-European-like” ancestry and examined across various medical specialties to assess differences in genomic variant classifications.

2. Statistical Approaches

The study employed two statistical methods: - Wilcoxon Signed-Rank Test: Paired genes were compared for allele prevalence differences between ancestry groups. - Chi-Square Test for Independence: Aimed to identify disparities in counts of unique variants exclusive to one population group.

Both approaches aimed to provide orthogonal validation and assess disparities in variant classifications across genetic ancestries.

3. Integration of MAVE Data

The study incorporated clinically calibrated MAVE data into the rules of the Clinical Genome Resource (ClinGen) Variant Curation Expert Panels (VCEP) for the reclassification of VUS in three widely studied genes: BRCA1, TP53, and PTEN.

4. Automated Reclassification Pipeline

An automated pipeline was developed to reclassify VUS based on ClinGen VCEP rules, with specific attention to using MAVE data for functional evidence codes. Essential codes were sequentially removed to identify their impact on reclassification outcomes.

Key Findings

1. Prevalence of VUS in Non-European Ancestry Populations

  • Non-European-like genetic ancestry groups displayed a statistically significant higher prevalence of VUS across all medical specialties and gene groups (p ≤ 5.95e−06) compared to individuals of European-like ancestry.
  • A greater prevalence of benign or likely benign (B/LB) and no clinical designation (ND) variants was also observed among non-European ancestry individuals (p ≤ 2.5e−05).
  • Conversely, pathogenic or likely pathogenic (P/LP) variants were more frequently classified in individuals of European-like ancestry (p ≤ 2.5e−05).

2. Reduction of VUS Disparity through MAVE Integration

  • Incorporating MAVE data significantly reduced VUS rates across non-European ancestry populations and showed a higher reclassification rate for VUS compared to European ancestry populations (p = 9.1e−03).
  • VUS reclassification rates for BRCA1, TP53, and PTEN:
    • European ancestry: 65.6% reclassified as “likely benign,” 0.8% as “benign,” and 3.3% as “likely pathogenic.”
    • Non-European ancestry: 67.1% reclassified as “likely benign,” 8.9% as “benign,” and 0.8% as “likely pathogenic.”

3. Impact of Evidence Code Inequities

  • MAVE evidence codes were equitably effective across ancestry groups, essential for the reclassification of 63.9–64.9% of VUS.
  • However, computational predictor evidence codes were more frequently essential for European-like individuals (49.8%) compared to non-European-like individuals (37.3%; p = 1.65e−03).
  • Similarly, allele frequency evidence codes disproportionately benefited European-like individuals (21.6%) over non-European-like populations (7.0%; p = 1.13e−05).

Conclusion

This study emphasizes the urgent need to prioritize saturation-style functional data production using MAVE as a solution to address health inequities in genomic medicine. Key findings demonstrate: 1. Higher baseline VUS rates in non-European ancestry populations largely result from insufficient understanding of their genetic diversity. 2. MAVE can counteract historical biases in computational tools and allele frequency databases by providing functional data for underrepresented variants. 3. Inequities in computational predictors and allele frequency usage underscore the necessity for unbiased training datasets that represent global genetic diversity.

Broader Implications

MAVE datasets not only mitigate existing disparities but also provide equitable training data for future computational predictors and artificial intelligence models. By offering a standardized platform to systematically catalog all variant effects, MAVE facilitates meaningful progress towards health equity.

Significance and Recommendations

  • The study highlights the disproportionate burden of VUS and the systemic barriers faced by non-European ancestry populations.
  • It underscores the need for expanding functional data efforts (e.g., MAVE) alongside increasing diversity in genomic recruitment.
  • Future research should investigate:
    • Expanding MAVE applications to address other genes and variant types (e.g., synonymous and in-frame indels).
    • Evaluating the disparities between Gain of Function (GOF) and Loss of Function (LOF) variations in clinical classifications.
    • Developing new computational predictors free from ancestry bias.

Research Contributions and Acknowledgments

This work was collaboratively supported by multiple funding agencies, including the Impact of Genomic Variation on Function (IGVF) consortia and the National Human Genome Research Institute (NHGRI). Authors acknowledged the contributions of All of Us participants and publicly available data from gnomAD for their analyses.

Efforts such as this study pave the way towards a more equitable genomic medicine landscape, ensuring all populations benefit from next-generation sequencing advancements in an inclusive and just manner.