TOP-LD: A tool to explore linkage disequilibrium with TOPMed whole-genome sequence data

Le Huang, Jonathan D. Rosen, Quan Sun, Jiawen Chen, Marsha M. Wheeler, Ying Zhou, Yuan‐I Min, Charles Kooperberg, Matthew P. Conomos, Adrienne M. Stilp, Stephen S. Rich, Jerome I. Rotter, Ani Manichaikul, Ruth J. F. Loos, Eimear E. Kenny, Thomas W. Blackwell, Albert V. Smith, Goo Jun, Fritz J. Sedlazeck, Ginger Metcalf, Eric Boerwinkle, Laura M. Raffield, Alex P. Reiner, Paul L. Auer, Yun LiThe American Journal of Human Genetics 2022 · DOI

TL;DR

Researchers have developed TOP-LD, a new online tool that provides significantly more comprehensive linkage disequilibrium (LD) information than existing tools. By leveraging high-coverage whole-genome sequencing data from the TOPMed program, TOP-LD offers a vastly expanded catalog of genetic variants, including rare ones and structural variants, crucial for advanced genetic analyses.

What Problem Does This Paper Address?

Understanding linkage disequilibrium (LD) – the non-random association of alleles at different genetic sites – is fundamental to many areas of genetics, from population studies to disease association research. Existing tools like HaploReg and LDlink rely on the 1000 Genomes Project data, which, while valuable, has limitations in its coverage of rare variants and the inclusion of structural variants. This gap means that researchers might miss important genetic signals or struggle to accurately fine-map associations, especially in diverse populations or when dealing with complex genetic variations. TOP-LD aims to bridge this gap by providing a more extensive and up-to-date LD reference.

Methods

The core of TOP-LD is its use of high-coverage (~30x) whole-genome sequencing (WGS) data from the NHLBI Trans-Omics for Precision Medicine (TOPMed) program. This data was processed to infer LD for four major ancestral populations: European (EUR), African (AFR), East Asian (EAS), and South Asian (SAS). The researchers specifically focused on genetically homogeneous groups by inferring local and global ancestry and filtering out related individuals. They then calculated LD for pairs of variants within 1 Mb, retaining those with an R² threshold of 0.2. A key novelty is the inclusion of tens of thousands of structural variants (SVs) alongside single-nucleotide variants (SNVs). The tool itself is a web-based application powered by MySQL, PHP, Javascript, and Apache, designed for fast batch queries.

Datasets

The primary dataset used is the NHLBI Trans-Omics for Precision Medicine (TOPMed) program's whole-genome sequencing data. Specifically, data from four cohorts (BioMe Biobank, MESA, JHS, and WHI) were utilized, comprising approximately 15,578 individuals. For comparison and validation, the paper references the 1000 Genomes Project data (Phase 1 and Phase 3), which is the basis for existing tools like HaploReg and LDlink.

Key Insights and Ideas

  • Vastly Expanded Variant Coverage: The most significant insight is the dramatic increase in the number of variants for which LD information is available. TOP-LD includes millions more variants than previous tools, particularly for rare alleles (MAF < 1%). This is crucial because many genetic associations are driven by these less common variants.
  • Inclusion of Structural Variants: The inclusion of tens of thousands of structural variants (SVs) is a major advancement. SVs, such as deletions and duplications, can have substantial functional impacts and were largely absent from previous LD reference panels.
  • Population-Specific LD: TOP-LD provides LD estimates tailored to specific ancestral populations, acknowledging that LD patterns can vary significantly between groups. This is essential for accurate genetic analyses in diverse cohorts.
  • Improved Fine-Mapping: The paper demonstrates that using TOP-LD as an LD reference for fine-mapping studies can identify more potentially causal variants compared to using the 1000 Genomes data. This is critical for pinpointing the exact genetic variants responsible for observed traits or diseases.

Key Results

  • Variant Coverage Increase: TOP-LD offers a 2.6- to 9.1-fold increase in variant coverage compared to HaploReg 4.0 or LDlink. For example, in the European population, TOP-LD includes 146.5 million autosomal SNVs, compared to 16.1 million in HaploReg 4.0/LDlink.
  • Rare Variant Focus: A large proportion of the additional variants in TOP-LD are low-frequency or rare. For instance, 137.8 million of the 146.5 million European autosomal SNVs have a MAF < 0.01.
  • High LD Proxy Availability: The majority of variants in TOP-LD have at least one LD proxy with a high R² threshold (≥ 0.8), indicating that even rare variants can often be tagged by more common ones.
  • Efficient Querying: The TOP-LD web interface is designed for speed, with single variant queries taking approximately 0.5 seconds and batch queries of 500 variants taking around 2.3 seconds.
  • Demonstrated Utility in Fine-Mapping: In a fine-mapping study of the GGT1 locus, TOP-LD-based credible sets identified three of seven "working truth" signals, compared to only one signal identified using the 1000 Genomes reference. When considering LD proxies, TOP-LD identified four signals versus one for the 1000 Genomes data.
  • Structural Variant Prioritization: The tool also aided in identifying a likely causal structural variant (a deletion) near the S1PR3 gene, which would have been missed by standard association analyses.

Limitations and Future Work

  • Population Specificity: While TOP-LD covers four major ancestral populations, it acknowledges that these are broad categories, and finer population stratification might be necessary for some applications.
  • Data Access: The data generated for this study is accessible via the TOP-LD web portal, but the underlying TOPMed WGS data itself has specific access requirements.
  • Future Directions: The authors highlight that TOP-LD can facilitate a wide range of applications beyond fine-mapping, including imputation of association summary statistics, construction of polygenic risk scores, and interpretation of GWAS results. The tool's comprehensive LD information is expected to be invaluable for future genetic research across diverse populations and variant types.

Summary updated

Abstract

Current publicly available tools that allow rapid exploration of linkage disequilibrium (LD) between markers (e.g., HaploReg and LDlink) are based on whole-genome sequence (WGS) data from 2,504 individuals in the 1000 Genomes Project. Here, we present TOP-LD, an online tool to explore LD inferred with high-coverage (∼30×) WGS data from 15,578 individuals in the NHLBI Trans-Omics for Precision Medicine (TOPMed) program. TOP-LD provides a significant upgrade compared to current LD tools, as the TOPMed WGS data provide a more comprehensive representation of genetic variation than the 1000 Genomes data, particularly for rare variants and in the specific populations that we analyzed. For example, TOP-LD encompasses LD information for 150.3, 62.2, and 36.7 million variants for European, African, and East Asian ancestral samples, respectively, offering 2.6- to 9.1-fold increase in variant coverage compared to HaploReg 4.0 or LDlink. In addition, TOP-LD includes tens of thousands of structural variants (SVs). We demonstrate the value of TOP-LD in fine-mapping at the GGT1 locus associated with gamma glutamyltransferase in the African ancestry participants in UK Biobank. Beyond fine-mapping, TOP-LD can facilitate a wide range of applications that are based on summary statistics and estimates of LD. TOP-LD is freely available online.
BibTeX
@article{huang2022top,
  author = {Le Huang and Jonathan D. Rosen and Quan Sun and Jiawen Chen and Marsha M. Wheeler and Ying Zhou and Yuan‐I Min and Charles Kooperberg and Matthew P. Conomos and Adrienne M. Stilp and Stephen S. Rich and Jerome I. Rotter and Ani Manichaikul and Ruth J. F. Loos and Eimear E. Kenny and Thomas W. Blackwell and Albert V. Smith and Goo Jun and Fritz J. Sedlazeck and Ginger Metcalf and Eric Boerwinkle and Laura M. Raffield and Alex P. Reiner and Paul L. Auer and Yun Li},
  title = {{TOP-LD: A tool to explore linkage disequilibrium with TOPMed whole-genome sequence data}},
  journal = {The American Journal of Human Genetics},
  year = {2022},
  doi = {10.1016/j.ajhg.2022.04.006},
}
Figure