The Human Pangenome’s sequence conservation reveals a landscape of polymorphic structural variations

HoJoon Lee, Stephanie Greer, Dmitri S. Pavlichin, Chris Hughes, Bo Zhou, Tsachy Weissman, Hanlee P. JibioRxiv 2022 · DOI

TL;DR

This paper introduces a novel k-mer indexing approach to analyze sequence conservation across the Human Pangenome, revealing a detailed landscape of structural variations. By identifying "pan-conserved segment tags" (PSTs), the researchers can precisely map polymorphic regions, offering a more comprehensive view of human genetic diversity than previous reference genomes.

What Problem Does This Paper Address?

The standard human genome reference (GRCh38) has limitations: it's a single linear representation, lacks representation of diverse populations, and has gaps. This makes it challenging to study complex genomic features like structural variations (SVs) that are crucial for understanding human disease and evolution. The Human Pangenome Project aims to address this by creating a more inclusive reference based on multiple high-quality haploid assemblies from diverse individuals. However, comparing these numerous assemblies to the existing reference efficiently is a significant computational challenge. This paper tackles this by developing a method to systematically identify and characterize conserved and divergent regions within the Pangenome, specifically focusing on structural variations.

Methods

The core of this paper's innovation lies in its k-mer indexing strategy. Instead of traditional sequence alignment, which can be computationally intensive for many genomes, they use short, fixed-length sequences (k-mers, specifically 31-mers) to index each genome assembly.

Here's a breakdown:

  • K-mer Indexing: Each genome assembly is broken down into overlapping 31-mers. These k-mers are then indexed by their genomic position and frequency.
  • Pan-Conserved Segment Tags (PSTs): The researchers identify k-mers that are present and unique in all assemblies. Consecutive unique k-mers are then merged to form "pan-conserved segment tags" (PSTs). These PSTs represent highly conserved regions across the Pangenome.
  • Interval Analysis: The key insight is that the intervals between these PSTs can reveal structural variations. If the length of an interval between two PSTs is consistent across assemblies, the region is considered conserved. If the length varies, it indicates a structural variation (like an insertion or deletion).
  • SV Visualization: A dot matrix plot approach using 31-mers is employed to visualize different types of SVs (insertions, deletions, duplications, inversions, and complex rearrangements) by comparing a reference genome (GRCh38) to a specific haploid assembly.

This k-mer approach is advantageous because it's scalable, can handle incomplete assemblies, and is efficient for comparing many genomes simultaneously.

Datasets

The study utilizes several key datasets:

  • Human Pangenome Reference Consortium (HPRC) Assemblies: 94 high-quality haploid assemblies from 47 individuals representing diverse populations (African, Admixed American, East Asian, South Asian, and Ashkenazim Jewish).
  • Reference Genomes: GRCh38 (the standard human reference) and CHM13 (a complete telomere-to-telomere human genome assembly).
  • Structural Variant (SV) Data:
    • Genome in a Bottle (GIAB) CMRG benchmark: Contains medically relevant SVs for benchmarking.
  • 1000 Genomes Project (1KGP) SVs: A large dataset of simple SVs from a diverse cohort.

Key Insights and Ideas

  • PSTs as Anchors: The concept of "pan-conserved segment tags" (PSTs) is a powerful tool. They act as reliable anchors across diverse genomes, allowing for precise measurement of genomic regions between them.
  • Interval Length as a Proxy for SVs: The variation in interval lengths between PSTs is a direct indicator of structural variation. This is a more sensitive and scalable approach than traditional SV calling methods for large-scale comparative genomics.
  • Pangenome Reveals Population-Specific Variations: The analysis highlights that structural variations are not uniformly distributed and can be specific to certain populations, underscoring the importance of a pangenome approach for capturing this diversity.
  • K-mer Indexing's Scalability: The k-mer indexing method proves highly effective for comparing a large number of genome assemblies, overcoming the computational bottlenecks of traditional alignment-based methods.

Key Results

  • Identification of 60,764 Polymorphic Intervals: The study identified a significant number of intervals with divergent lengths, pointing to structural polymorphisms across the Pangenome.
  • High Conservation: Approximately 83.6% of the analyzed intervals were found to be "conserved" (identical sequence lengths) across all 94 HPRC assemblies.
  • Overlap with Known SVs: A high percentage (93.8%) of simple SVs from the 1000 Genomes Project overlapped with the identified polymorphic intervals, validating the approach.
  • Characterization of SV Types: The SV plots successfully visualized various SV types, including insertions, deletions, tandem duplications, multiple duplications, inversions, and complex rearrangements.
  • Population-Specific Divergence: The analysis revealed population-specific divergent intervals, with some variations found exclusively in certain superpopulations.

Limitations and Future Work

  • PST Mapping Limitations: PSTs have limitations when mapped to contigs, especially near telomeres. The Pangenome's current assembly structure (not yet complete chromosomes) also creates breaks that can prevent some tandem segments from being compared.
  • Computational Tools for Breakpoints: The authors plan to develop computational tools to precisely determine SV breakpoints using k-mer plots and PSTs.
  • Expanding the Pangenome: The Human Pangenome Reference Consortium plans to expand the reference to include more haploid assemblies, which will allow for more comprehensive updates and analyses.
  • Resource for the Community: The study provides an interval matrix in BED format, making it accessible for researchers to explore regions of interest within the Pangenome.

Summary updated

Abstract

ABSTRACT The Human Pangenome is a new reference build that addresses many of the limitations of the current reference. The first release is based on 94 high-quality haploid assemblies from individuals with diverse backgrounds. To facilitate the Pangenomes adoption in the wider research community, we conducted a multiple genome comparative analysis against the current GRCh38 reference. We applied a k-mer indexing strategy to identify highly conserved sequences that are omnipresent across all Pangenome assemblies, the reference and CHM13, a reference assembly with telomere-to-telomere chromosomes. Pan-conserved tag segments provided an informative set of universally conserved sequences. Examining the intervals between pairs of these segments defined highly conserved segments of the genome versus ones that have structurally related polymorphisms. We identified a Pangenome landscape of 60,764 polymorphic intervals with unique and geo-ethnic features. Overall, this study of the Pangenome revealed the conserved versus divergent features including the landscape of polymorphic structural variants.
BibTeX
@misc{lee2022human,
  author = {HoJoon Lee and Stephanie Greer and Dmitri S. Pavlichin and Chris Hughes and Bo Zhou and Tsachy Weissman and Hanlee P. Ji},
  title = {{The Human Pangenome’s sequence conservation reveals a landscape of polymorphic structural variations}},
  year = {2022},
  doi = {10.1101/2022.10.06.511239},
  howpublished = {bioRxiv},
}
Figure