Defining a tandem repeat catalog and variation clusters for genome-wide analyses

This paper introduces the TRExplorer catalog v1.0, a comprehensive, 4.86-million-locus tandem repeat resource designed for both short- and long-read analyses that resolves inconsistencies in existing catalogs by defining variation clusters through a novel algorithm, thereby enabling more accurate genome-wide genotyping and facilitating a unified portal for data access.

The Gist

Imagine your genome is a massive, ancient library containing the instructions for building a human. Most of the books in this library are written in a clear, standard alphabet (the A, C, G, and T letters of DNA). However, there are also thousands of pages filled with repeating patterns, like "CAG CAG CAG CAG" or "AAAAA." These are called Tandem Repeats (TRs).

For a long time, scientists have been trying to count these repeats to understand diseases, but they've been using different, conflicting maps to find them. Some maps say a repeating section starts here; others say it starts there. It's like trying to measure a room where one person counts from the left wall and another counts from the right wall, leading to different results and confusion.

This paper introduces a new, universal map called TRExplorer to fix this mess. Here is the breakdown of what they did, using simple analogies:

1. The Problem: Too Many Different Maps

Before this paper, researchers used various "catalogs" (lists of where these repeats are).

• The Conflict: One catalog might say a repeat is 10 letters long, while another says it's 12.
• The Result: If two scientists study the same disease using different catalogs, they get different answers. It's like two chefs trying to bake the same cake but using different recipes; one adds salt, the other adds sugar, and the cakes turn out totally different.
• The Gap: Old maps were missing many repeats, especially the messy ones that look different from person to person.

2. The Solution: The "TRExplorer" Master Map

The authors built a new, massive catalog called TRExplorer v1.0. Think of this as the "Google Maps" of tandem repeats.

• Size: It contains nearly 5 million repeat locations.
• Completeness: It combines the best parts of old maps (which are good for simple, clean repeats) with new data from long-read sequencing (which can see through the messy, complex repeats).
• Compatibility: It is designed to work with both old, cheap technology (short-read sequencing) and new, expensive technology (long-read sequencing), so everyone can use it.

3. The Big Innovation: "Variation Clusters"

This is the most creative part of the paper. The authors realized that some repeats don't exist in isolation. They are often surrounded by a chaotic neighborhood of other mutations.

• The Analogy: Imagine a single tree (a repeat) in a forest.
  • Isolated Tree: Sometimes, the tree stands alone in a clear field. You can easily count its rings.
  • The "Variation Cluster": Sometimes, the tree is in a dense, tangled thicket where vines, rocks, and other trees are shifting and moving. If you try to count just that one tree, you might get it wrong because the ground around it is unstable.

The authors created a new tool called vclust (Variation Cluster) that doesn't just look at the single tree. Instead, it draws a boundary around the entire thicket.

• Why this matters: Instead of trying to count the rings of one tree in a storm, they analyze the whole storm. This allows them to see the entire picture of the genetic variation, leading to much more accurate diagnoses. They found over 25,000 of these complex "thickets" in the human genome.

4. The "Treasure Hunt" Portal

To make this useful for everyone, they built a free website (trexplorer.broadinstitute.org).

• Think of this as a search engine for your genetic library.
• You can search for specific types of repeats (e.g., "Show me all the CAG repeats in the brain").
• You can filter them (e.g., "Only show me the ones that vary between people").
• You can download the data to use in your own research.

5. Why This Changes Everything

• For Doctors: It reduces the risk of misdiagnosis. If a doctor uses the new map, they won't accidentally miss a dangerous repeat or miscount it because of a confusing boundary.
• For Scientists: It stops the "Tower of Babel" problem. Now, a scientist in Australia and a scientist in the US can use the exact same map, ensuring their results match perfectly.
• For the Future: As we learn more about how these repeats cause diseases (like Huntington's or ALS), having a single, high-quality map ensures we don't waste time arguing over where the repeat actually starts and ends.

In a nutshell:
The authors built a universal, high-definition atlas for the repetitive parts of our DNA. They realized that some repeats are messy neighborhoods, not just single houses, and they created a new way to map those neighborhoods. This tool will help doctors and scientists finally speak the same language when studying genetic diseases caused by these repeating patterns.

Technical Summary

1. Problem Statement

Tandem repeats (TRs), including Short Tandem Repeats (STRs) and Variable Number Tandem Repeats (VNTRs), are critical genomic features associated with mutation rates, gene regulation, and numerous diseases. However, genome-wide TR analysis faces significant challenges due to a lack of consensus in TR catalog design:

• Inconsistency: Existing catalogs (ranging from ~200k to >7M loci) often disagree on locus boundaries and definitions. Even minor differences in boundary definitions can lead to incorrect pathogenic thresholds (e.g., the PABPN1 locus for Oculopharyngeal Muscular Dystrophy).
• Fragmentation: Different catalogs are optimized for specific technologies (short-read vs. long-read) or strategies (reference-based vs. cohort-driven), leading to incompatibility across studies and resources.
• Complexity: Many TRs reside in complex regions with surrounding polymorphisms (interruptions, indels, nested repeats). Standard genotyping tools often fail in these regions, producing errors when analyzing repeats individually rather than as part of a larger variable region.
• Missing Data: Reference-based catalogs miss polymorphic repeats absent from the reference genome, while cohort-driven catalogs may lack comprehensive boundary definitions suitable for short-read analysis.

2. Methodology

The authors developed a unified, comprehensive catalog and a novel analytical framework to address these issues.

A. Construction of the TRExplorer Catalog v1.0

The catalog was built by integrating four distinct source catalogs to maximize coverage and compatibility:

1. Disease-Associated Loci: 63 known pathogenic loci and their neighbors, prioritizing established definitions for clinical compatibility.
2. Illumina Polymorphic Catalog: 174,244 loci derived from 1000 Genomes Project short-read data, ensuring compatibility with existing short-read tools.
3. Reference-Based Perfect Repeats: 4.39 million loci identified by scanning the hg38 reference for perfect repeats (≥3 copies of motifs 3–1000 bp, or specific thresholds for homopolymers/dinucleotides).
4. Cohort-Driven Polymorphic Loci: 297,519 loci detected from 78 haplotype-resolved T2T (Telomere-to-Telomere) assemblies, capturing repeats with fewer than 3 copies in the reference or those with single-base interruptions.

Key Features of the Catalog:

• Total Size: 4.86 million TR loci (4.8M STRs, ~60k VNTRs).
• Stratification: Loci are categorized into:
  • Isolated TRs: Repeats surrounded by non-polymorphic flanking regions (suitable for copy-number analysis).
  • Variation Clusters (VCs): Regions where TRs are embedded in broader polymorphic sequences requiring sequence-level analysis.
• Format: Provided in formats compatible with both short-read (e.g., ExpansionHunter, HipSTR) and long-read (e.g., TRGT, LongTR) genotyping tools.

B. The vclust Algorithm for Variation Clusters

To define Variation Clusters, the authors introduced vclust, a novel algorithm leveraging long-read HiFi sequencing data:

• Mechanism: It calculates the probability that bases flanking a TR belong to a variation cluster based on local polymorphism patterns (mismatches, indels, soft-clips) observed in aligned reads.
• Process: It iteratively extends the boundaries of a TR upstream and downstream until the probability of the flanking base being part of the cluster drops below 0.5.
• Output: Defines a contiguous genomic region containing the TR and its surrounding polymorphic context.

C. Validation and Analysis

• Polymorphism Assessment: Genotyped 1,283 HiFi samples (HPRC and All of Us) to determine allele frequency distributions and standard deviations.
• Accuracy Testing: Compared genotyping accuracy of individual TRs vs. full Variation Clusters against high-quality T2T assemblies (HG002).
• Portal Development: Created trexplorer.broadinstitute.org for interactive search, visualization, and export of the catalog.

3. Key Results

Catalog Statistics

• Composition: 4,803,366 STRs and 59,675 VNTRs.
• Novelty: Contains ~780k STRs and ~22k VNTRs that are polymorphic and absent from widely used short-read catalogs.
• Genomic Distribution: 65.8% intronic, 27.8% intergenic, 4.1% exonic/promoter/UTR.
• Polymorphism: 52% of loci are polymorphic (standard deviation ≥ 0.2) in the tested HiFi datasets. Trinucleotide repeats showed the highest fraction of non-polymorphic loci.

Variation Clusters (VCs)

• Discovery: Identified 273,112 Variation Clusters covering 744,458 TRs.
• Complexity:
  • 66% of VCs overlap two or more TRs.
  • 18% of VCs span >120 bp (a length where short-read genotyping becomes difficult).
  • Complex VCs: 24,867 VCs contain ≥5 constituent TRs. These have a median length of 208 bp, with 86% exceeding 120 bp.
• Accuracy Improvement: Genotyping entire VCs significantly improved sequence concordance compared to genotyping individual TRs:
  • Isolated TRs: 96.21% concordance.
  • TRs within VCs (individual): 79.43% concordance.
  • Full VCs: 90.58% concordance (rising to 95.98% if off-by-one errors are ignored).
• Feature Correlation: TRs within VCs are more likely to be in intronic/intergenic regions, have lower mappability, and be associated with complex repeat motifs (>10 bp).

Cost and Utility

• Cost Analysis: Genotyping the full 4.86M catalog costs several dollars per sample. However, filtering for specific subsets (e.g., disease-associated motifs or highly polymorphic loci) reduces costs to $0.02–$0.20 per sample, making large-scale population studies feasible.
• Tool Compatibility: The catalog successfully bridges the gap between short-read and long-read analyses, allowing researchers to select appropriate subsets based on their sequencing technology.

4. Significance and Impact

• Standardization: Provides the first consensus-based, comprehensive TR catalog designed for both short-read and long-read technologies, addressing the fragmentation of previous resources.
• Clinical Relevance: By standardizing locus boundaries (e.g., for PABPN1, C9orf72), the catalog reduces the risk of diagnostic errors in genetic testing.
• Methodological Advancement: The introduction of Variation Clusters shifts the paradigm from analyzing repeats in isolation to analyzing them within their polymorphic context. This is crucial for resolving complex regions where standard genotyping fails.
• Community Resource: The open-access TRExplorer portal allows researchers to filter, visualize, and download tailored subsets of the catalog, facilitating reproducible and comparable TR studies across the global research community.
• Future-Proofing: The catalog serves as a foundation for future population-scale TR studies, enabling the integration of methylation data, somatic mosaicism, and diverse population allele frequencies.

In conclusion, this work establishes a critical infrastructure for the next generation of tandem repeat genomics, moving the field from fragmented, tool-specific catalogs toward a unified, high-resolution resource capable of supporting both clinical diagnostics and large-scale population genetics.