Merqury: reference-free quality, completeness, and phasing assessment for genome assemblies.

Rhie A, Walenz BP, Koren S, Phillippy AM. · 2020

Read Merqury: reference-free quality, completeness, and phasing assessment for genome assemblies. by Rhie A, Walenz BP, Koren S, Phillippy AM. (2020) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc7488777.txt.

Introduction

Rhie, Walenz, Koren, and Phillippy present Merqury, a tool that grades genome assemblies without a reference. New long-read assemblies can beat the references on hand. That makes standard checks hard to run.

Merqury compares short sequence words, called k-mers, in an assembly with those in accurate raw reads. From that match it estimates base accuracy and completeness. For family trios, it also scores each parental copy on its own. It can check phasing blocks and switch errors too. Plots of the k-mer spectrum help users see problems at a glance. Tests on human and plant genomes showed the tool is fast and robust. It gives assembly teams a shared yardstick for quality claims.

Catalog record

CREATORRhie A, Walenz BP, Koren S, Phillippy AM.
YEAR2020
COLLECTIONPAPERS
FORMATTXT
FILE SIZE76 KB
LICENSECC BY 4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/articles/PMC7488777

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Merqury: reference-free quality, completeness, and phasing assessment for genome assemblies By Arang Rhie, Brian P Walenz, Sergey Koren, Adam M Phillippy Published in: Genome Biology DOI: 10.1186/s13059-020-02134-9 ABSTRACT Abstract Recent long-read assemblies often exceed the quality and completeness of available reference genomes, making validation challenging. Here we present Merqury, a novel tool for reference-free assembly evaluation based on efficient k-mer set operations. By comparing k-mers in a de novo assembly to those found in unassembled high-accuracy reads, Merqury estimates base-level accuracy and completeness. For trios, Merqury can also evaluate haplotype-specific accuracy, completeness, phase block continuity, and switch errors. Multiple visualizations, such as k-mer spectrum plots, can be generated for evaluation. We demonstrate on both human and plant genomes that Merqury is a fast and robust method for assembly validation. Keywords: Genome assembly, Assembly validation, Benchmarking, K-mers, Haplotype phasing, Trio binning BACKGROUND With recent advances in long-read [1–3] and long-range sequencing technologies [4–6], new assembly pipelines are generating more continuous, complete, and accurate diploid genome assemblies than ever before [4, 7–14]. However, de novo assembled genomes are difficult to validate due to the lack of a known truth. Existing methods use Illumina reads to infer base-level accuracy by aligning the reads to the assembly for evaluation [15]. Base errors in the consensus are detected as variants (SNPs or small indels) when aligning the short reads. However, this method is heavily reliant on the short-read mapping, which could be biased in repetitive regions, under-collapsed regions, or regions of low consensus accuracy. For measuring completeness and false duplications, near-universal single-copy orthologs (BUSCOs) [16] have been widely used to evaluate the gene content of assemblies. BUSCO is robust for species that have been widely studied, such as human and mouse. However, this analysis can be inaccurate when the newly assembled genome contains true copy number or sequence variants that were not considered when building the initial BUSCO gene set. In addition, BUSCO only examines conserved single-copy genes and fails to evaluate the most difficult-to-assemble regions of the genome. In contrast, k-mers (genomic substrings of length k) can be used in a reference-free manner for assessing genome assembly quality metrics. Genome assembly validation via k-mer copy number analysis was introduced by Mapleson et al. in their KAT tool [17], which enables visual inspection of k-mer spectra plots to identify artificial duplications and missing sequences. Merqury takes much of its inspiration from the ideas introduced by KAT. QUAST-LG [18] is another assembly validation tool that provides both BUSCO and reference-based k-mer measurements, as well as alignment metrics against a closely related reference genome. However, many of QUAST’s metrics are based on the reference alignment, which incorrectly reports true variants in the assembled genome as potential mis-assemblies. A recent extension to the QUAST toolkit combines evidence from both k-mers and long reads to better identify certain classes of structural mis-assembly [19]. Assessing haplotype phasing without a truth set is also challenging. Diploid genome assemblers generate both primary and alternate assemblies representing the two haplotypes. The primary assembly is typically a pseudo-haplotype that captures both the homozygous regions along with a single copy of the heterozygous alleles. Such a pseudo-haplotype does not guarantee long-range phasing, so to estimate phase block statistics, the alternate alleles must be mapped back to the primary assembly to determine regions corresponding to the primary-alternate haplotype phase blocks [9]. However, this can be challenging when the alternate alleles do not map well to the primary sequence due to high sequence divergence or mis-assemblies. Moreover, long-read assemblies often collapse regions of low heterozygosity, which are excluded when calculating phase block statistics, thus over-representing the correctness. Alternative methods report phasing statistics from small variants (mostly SNPs) called with short-read mapping [8, 20–23], or use benchmark genomes that have curated, phased variation call sets [24–27]. Both methods rely on a reference sequence as the primary source to detect heterozygous variations. However, these reference sequences are incomplete, and recent studies have demonstrated the shortcomings of the current human reference genome and variant call sets [8, 21, 28]. For example, the highly variable major histocompatibility complex (MHC) is excluded from the Genome in a Bottle (GIAB) and Global Alliance for Genomic Health (GA4GH) reference panels [26] due to its repetitive nature and need of a specialized mapping strategy to account for the high allelic diversity [29]. Moreover, reference-guided strategies require significant manual curation and effort and will not scale as large cohort sequencing projects become common [30, 31]. This validation strategy is also not applicable to any species without a curated and complete reference [32]. To overcome these limitations, we developed Merqury, which generates assembly assessment metrics using k-mers alone. Merqury compares a set of k-mers derived from unassembled, high-accuracy sequencing reads to a genome assembly for evaluation. Merqury builds upon the k-mer-based analyses introduced by KAT but adds new functionality for evaluating the accuracy of phased diploid genome assemblies (Table 1). Unique to Merqury, the generated metrics include consensus quality (QV) and k-mer completeness, and when parental genomic sequences are available (either assembled or unassembled), Merqury can output haplotype completeness, phase block statistics, switch error rates, and visual representations of phase consistency for the child’s genome. This includes TDF (or BED) features that can be displayed in a genome browser for visualizing the presence of k-mer classes across a genome (e.g., the k-mers inherited from a parental genome). We show that Merqury’s k-mer-based assembly validation produces comparable or better results than existing methods, such as BUSCO gene completeness and mapping-based measurements. [Table: Feature comparison of KAT and Merqury] RESULTS To demonstrate the ability of Merqury to evaluate the accuracy, completeness, and phasing of an assembly, we first applied it to an Arabidopsis thaliana F1 hybrid [9], for which the parental strains (Col-0 and Cvi-0, simplified as Col and Cvi) have also been sequenced. For a comparison of multiple assemblies, we demonstrate Merqury on haplotype-resolved (TrioCanu [10]), pseudo-haplotype (FALCON-Unzip [9]), and mixed-haplotype (Canu [33]) assemblies of this hybrid genome. Total assembly size is typically used as a rough measure of haplotype completeness. For example, the TrioCanu haplotype assemblies have similar total bases, 122~124 Mbp (Table 2), close to the expected haploid genome size of 130 Mbp, indicating the haplotype assemblies are well balanced (assuming haplotype-specific bases are evenly inherited). In comparison, the primary assembly of FALCON-Unzip has ~ 35 Mbp bases more than the alternate assembly. However, it is difficult to understand where this difference originates from the assembly size alone. The mixed-haplotype Canu assembly, in comparison, is ~ 100 Mbp larger than the expected genome size of 130 Mb. Again, we can assume this assembly resolved both haplotypes, but since the haplotypes have been combined in a single assembly, we cannot know the composition from the size and continuity measures alone. In the following sections, we describe how Merqury’s statistics and plots can be used to dissect and understand these assemblies. [Table: Merqury quality and completeness statistics for Arabidopsis thaliana and human genome assemblies] Copy number spectrum We counted k-mers from Illumina whole-genome sequencing of the A. thaliana F1 hybrid, as well as from each assembly, using Meryl, a k-mer counting tool we extended to support k-mer set operations for Merqury (Methods). The copy number spectrum plot, known

Read the complete text