Third-generation sequencing has revolutionised de novo genome assembly. Thanks to this technology, genome reference sequences have evolved from fragmented drafts to gapless, telomere-to-telomere genome assemblies. Long reads produced by Oxford Nanopore and PacBio technologies can span structural variants and resolve complex repetitive regions such as centromeres, unlocking previously inaccessible genomic information. Nowadays, many research groups can afford to sequence the genome of their working model using long reads. Nevertheless, genome assembly poses a significant computational challenge. Read length, quality, coverage and genomic features such as repeat content can affect assembly contiguity, accuracy, and completeness in almost unpredictable ways. Consequently, there is no best universal software or protocol for this task. Producing a high-quality assembly requires chaining several tools into pipelines and performing extensive comparisons between the assemblies obtained by different tool combinations to decide which one is the best. This task can be extremely challenging, as the number of tools available rises very rapidly, and thorough benchmarks cannot be updated and published at such a fast pace. 

In their paper, Orjuela and collaborators present CulebrONT [1], a universal pipeline that greatly contributes to overcoming these challenges and facilitates long-read genome assembly for all taxonomic groups. CulebrONT incorporates six commonly used assemblers and allows to perform assembly, circularization (if needed), polishing, and evaluation in a simple framework. One important aspect of CulebrONT is its modularity, which allows the activation or deactivation of specific tools, giving great flexibility to the user. Nevertheless, possibly the best feature of CulebrONT is the opportunity to benchmark the selected tool combinations based on the excellent report generated by the pipeline. This HTML report aggregates the output of several tools for quality evaluation of the assemblies (e.g. BUSCO [2] or QUAST [3]) generated by the different assemblers, in addition to the running time and configuration parameters. Such information is of great help to identify the best-suited pipeline, as exemplified by the authors using four datasets of different taxonomic origins. Finally, CulebrONT can handle multiple samples in parallel, which makes it a good solution for laboratories looking for multiple assemblies on a large scale. 

References

1. Orjuela J, Comte A, Ravel S, Charriat F, Vi T, Sabot F, Cunnac S (2022) CulebrONT: a streamlined long reads multi-assembler pipeline for prokaryotic and eukaryotic genomes. bioRxiv, 2021.07.19.452922, ver. 5 peer-reviewed and recommended by Peer Community in Genomics. https://doi.org/10.1101/2021.07.19.452922

2. Simão FA, Waterhouse RM, Ioannidis P, Kriventseva EV, Zdobnov EM (2015) BUSCO: assessing genome assembly and annotation completeness with single-copy orthologs. Bioinformatics, 31, 3210–3212. https://doi.org/10.1093/bioinformatics/btv351

3. Gurevich A, Saveliev V, Vyahhi N, Tesler G (2013) QUAST: quality assessment tool for genome assemblies. Bioinformatics, 29, 1072–1075. https://doi.org/10.1093/bioinformatics/btt086

In their paper, Guiglielmoni et al. propose we pick up our snorkels and palms and take "A deep dive into genome assemblies of non-vertebrate animals" (1). Indeed, while numerous assembly-related tools were developed and tested for human genomes (or at least vertebrates such as mice), very few were tested on non-vertebrate animals so far. Moreover, most of the benchmarks are aimed at raw assembly tools, and very few offer a guide from raw reads to an almost finished assembly, including quality control and phasing.

This huge and exhaustive review starts with an overview of the current sequencing technologies, followed by the theory of the different approaches for assembly and their implementation. For each approach, the authors present some of the most representative tools, as well as the limits of the approach.

The authors additionally present all the steps required to obtain an almost complete assembly at a chromosome-scale, with all the different technologies currently available for scaffolding, QC, and phasing, and the way these tools can be applied to non-vertebrates animals. Finally, they propose some useful advice on the choice of the different approaches (but not always tools, see below), and advocate for a robust genome database with all information on the way the assembly was obtained.

This review is a very complete one for now and is a very good starting point for any student or scientist interested to start working on genome assembly, from either model or non-model organisms. However, the authors do not provide a list of tools or a benchmark of them as a recommendation. Why? Because such a proposal may be obsolete in less than a year.... Indeed, with the explosion of the 3rd generation of sequencing technology, assembly tools (from different steps) are constantly evolving, and their relative performance increases on a monthly basis. In addition, some tools are really efficient at the time of a review or of an article, but are not further developed later on, and thus will not evolve with the technology. We have all seen it with wonderful tools such as Chiron (2) or TopHat (3), which were very promising ones, but cannot be developed further due to the stop of the project, the end of the contract of the post-doc in charge of the development, or the decision of the developer to switch to another paradigm. Such advice would, therefore, need to be constantly updated.

Thus, the manuscript from Guiglielmoni et al will be an almost intemporal one (up to the next sequencing revolution at last), and as they advocated for a more informed genome database, I think we should consider a rolling benchmarking system (tools, genome and sequence dataset) allowing to keep the performance of the tools up-to-date, and to propose the best set of assembly tools for a given type of genome.

References

1. Guiglielmoni N, Rivera-Vicéns R, Koszul R, Flot J-F (2022) A Deep Dive into Genome Assemblies of Non-vertebrate Animals. Preprints, 2021110170, ver. 3 peer-reviewed and recommended by Peer Community in Genomics. https://doi.org/10.20944/preprints202111.0170

2. Teng H, Cao MD, Hall MB, Duarte T, Wang S, Coin LJM (2018) Chiron: translating nanopore raw signal directly into nucleotide sequence using deep learning. GigaScience, 7, giy037. https://doi.org/10.1093/gigascience/giy037

3. Trapnell C, Pachter L, Salzberg SL (2009) TopHat: discovering splice junctions with RNA-Seq. Bioinformatics, 25, 1105–1111. https://doi.org/10.1093/bioinformatics/btp120