Significantly improving the quality of genome assemblies through curation.

Howe K, Chow W, Collins J, Pelan S, Pointon DL, Sims Y, Torrance J, Tracey A, Wood J. · 2021

Read Significantly improving the quality of genome assemblies through curation. by Howe K, Chow W, Collins J, Pelan S, Pointon DL, Sims Y, Torrance J, Tracey A, Wood J. (2021) 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-pmc7794651.txt.

Introduction

Howe and co-authors explain how manual curation greatly improves genome assemblies. Assemblies underpin much of modern biology. Error-free assemblies remain an unmet goal. Automated pipelines still miss that mark for complex genomes.

Expert review bridges the gap for now. Curators check assemblies and fix errors by hand. This cuts error counts a lot. Lessons from curation feed back into the automated pipelines. The authors describe their method with gEVAL, a genome evaluation browser. They also give advice that works without gEVAL. The goal is wider use of curation across the genomics community.

Catalog record

CREATORHowe K, Chow W, Collins J, Pelan S, Pointon DL, Sims Y, Torrance J, Tracey A, Wood J.
YEAR2021
COLLECTIONPAPERS
FORMATTXT
FILE SIZE35 KB
LICENSECC BY 4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/articles/PMC7794651

Browse the full library · About the Tower of Babel

Full text

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

Significantly improving the quality of genome assemblies through curation By Kerstin Howe, William Chow, Joanna Collins, Sarah Pelan, Damon-Lee Pointon, Ying Sims, James Torrance, Alan Tracey, Jonathan Wood Published in: GigaScience DOI: 10.1093/gigascience/giaa153 ABSTRACT Abstract Genome sequence assemblies provide the basis for our understanding of biology. Generating error-free assemblies is therefore the ultimate, but sadly still unachieved goal of a multitude of research projects. Despite the ever-advancing improvements in data generation, assembly algorithms and pipelines, no automated approach has so far reliably generated near error-free genome assemblies for eukaryotes. Whilst working towards improved datasets and fully automated pipelines, assembly evaluation and curation is actively used to bridge this shortcoming and significantly reduce the number of assembly errors. In addition to this increase in product value, the insights gained from assembly curation are fed back into the automated assembly strategy and contribute to notable improvements in genome assembly quality. We describe our tried and tested approach for assembly curation using gEVAL, the genome evaluation browser. We outline the procedures applied to genome curation using gEVAL and also our recommendations for assembly curation in a gEVAL-independent context to facilitate the uptake of genome curation in the wider community. Keywords: genome, assembly, curation, gEVAL ASSEMBLY CURATION ADDS SIGNIFICANT VALUE Despite the advances in sequencing and mapping technologies and the ever-increasing number of sophisticated algorithms and pipelines available, generating error-free eukaryotic genome assemblies in a purely automated fashion is currently not possible [1, 2]. Assembly software designed to generate continuous sequence from raw reads is confused by heterozygous or repeat-rich regions, introducing erroneous duplications, collapses, and misjoins. The same issues recur in subsequent scaffolding processes that aim to turn primary contigs into representations of chromosomal units. The fact that these tools are commonly applied in series rather than in parallel results in the passing of mistakes made from one process on to the next. As a result, even so-called high-quality or “platinum” assemblies can suffer from hundreds to thousands of duplications, collapses, misjoins and missed joins. Because assemblies are often judged simply by their continuity rather than by their completeness and (structural) correctness, these errors go unnoticed. This affects research in many ways, making whole regions of the genome impossible to access or misleading researchers who misinterpret assembly artefacts as biological findings (C. Lee, unpublished data). One way to address these shortcomings is in-depth analysis of discordances between the assembly that has been generated and the different data types available for the sequenced individual or species and subsequent resolution of these discordances. This can be performed at the sequence and the structural level. Many automated tools are available that assess sequence quality through read alignment, k-mer counting, gene finding, and other methods [3–6]. For structural quality assessment, several individual tools can be used, but these tend to analyse a single data type at a time rather than combining insights from analysis of several in parallel [7, 8]. We created gEVAL, the genome evaluation browser, to enable a user to visualize and evaluate discordances between an assembly and multiple sets of accompanying data at the same time [9]. gEVAL enables the identification of errors and simultaneously suggests ways to resolve them. Combined with manual assessment of the generated data by experienced curators and a pipeline that enables the curators to record changes and recreate the improved assembly accordingly, gEVAL provides a critical addition to strategies striving to produce assemblies of the highest possible quality. Herein we outline the strategic design, achievements, and limitations of the gEVAL approach to assembly curation. gEVAL is tied into our local infrastructure and as such sadly not portable, yet fully publicly accessible at geval.org.uk. We therefore also provide detailed recommendations on how to create similar analyses that do not use gEVAL to promote the core, proven design concepts in gEVAL. This is especially timely in the context of emerging projects that aim to assemble the genomes of very large numbers of species to highest quality possible, including the Vertebrate Genomes Project (VGP), the Darwin Tree of Life Project (DToL, darwintreeoflife.org), and the overarching Earth Biogenome Project [1, 10]. CHECKING FOR ASSEMBLY COHERENCE, COVERAGE, AND CONTAMINATION We recommend that every genome assembly be checked for coherence. This includes making sure that only data that belong to the relevant species are used for assembly in the first place. This is best checked before starting the assembly process by aligning all raw datasets with, e.g., mash [11] and checking that the data are in fact combinable (i.e., that they are likely to derive from the same underlying distribution of sequence). A major source of remaining technical error in assemblies is the retention of duplicated regions that result from failure to recognize that two sequences are in fact allelic. These false duplications have wide-ranging negative consequences for subsequent research, e.g., causing prediction of erroneous gene duplications [1]. False duplications are caused by either incorrect resolution of assembly graphs or failures in detection of haplotypic variation. They can be detected using simple read coverage plots or more sophisticated k-mer analyses (e.g., using the K-mer Analysis Toolkit [KAT] [4], the K-mer Counter [KMC] [12], or Merqury [6]). K-mer approaches also support the estimation of the completeness of the assembly (i.e., whether the assembly contains all the relevant k-mers present in the reads) and the ploidy of the genome [13]. False duplications can be removed, ideally after generation of the contigs, with tools that recognize partial and complete allele overlap, such as purge_dups [14]. In addition to duplications, assembly quality is also negatively affected by erroneous sequence collapses, mostly located in repetitive regions. Collapses are relatively easy to detect on the basis of increased read coverage, but harder to resolve because they require generation of new sequence. This can be performed through extraction of mapped reads and local reassembly under more stringent conditions, or with more sophisticated methods such as the Segmental Duplication Assembler (SDA) [15]. Assemblies are frequently polished after contig generation, using the bulk of data or particular high base accuracy data such as Illumina short reads, to correct remaining errors in the derived consensus sequence. It is however possible to over-polish, such that rare repeat variants are replaced by the most abundant version, or where nuclear insertions of organellar genome fragments (nuclear mitochondrial transfers [NUMTs] and nuclear plastid transfers [NUPTs]) are polished to match the organelle sequence. For polishing, the target genome assembly therefore must include the organelle genomes. Organellar genomes are often missing from assemblies because assembly toolkits recognize and exclude them as repeat sequence or because they yield complex graphs that conflict with nuclear insertions. They can be assembled independently from the raw reads, e.g., using the mitoVGP pipeline [16]. Contigs/scaffolds that represent the organelle genomes should be identified and submitted as such to the International Nucleotide Sequence Database Collaboration (INSDC) archives. A preliminary assembly of data from a target species can inadvertently include synthetic sequence from cloning or sequencing systems, contamination from species handled in the same laboratory or sequencing centre, or contamination from natural cobionts of the target (e.g., gut and skin microbiomes, unsuspected parasites). Decontamination serves to detect and mask or remove sequence not originating from the target species, and to separate organelle genomes from the primary assembly if

Read the complete text