compleasm: a faster and more accurate reimplementation of BUSCO.
Huang N, Li H. · 2023
Read compleasm: a faster and more accurate reimplementation of BUSCO. by Huang N, Li H. (2023) 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-pmc10558035.txt.
Introduction
Huang and Li built compleasm in 2023, a faster tool to grade how complete a genome build is. A genome build may miss genes, which then warps all later work. The standard grader, BUSCO, checks for a set of genes kept near unchanged across species. It runs slowly, most of all on huge genomes.
Their tool pairs a fast protein aligner with the same gene sets. Huang reports a 14-fold speed gain on human builds. It also scored the full human build at 99.6% complete, close to the 99.5% from gene labels. BUSCO had scored the same build at just 95.7%. The code is free on GitHub for any lab to use.
Catalog record
| CREATOR | Huang N, Li H. |
|---|---|
| YEAR | 2023 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 20 KB |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
compleasm: a faster and more accurate reimplementation of BUSCO By Neng Huang, Heng Li, Tobias Marschall Published in: Bioinformatics DOI: 10.1093/bioinformatics/btad595 ABSTRACT Abstract Motivation Evaluating the gene completeness is critical to measuring the quality of a genome assembly. An incomplete assembly can lead to errors in gene predictions, annotation, and other downstream analyses. Benchmarking Universal Single-Copy Orthologs (BUSCO) is a widely used tool for assessing the completeness of genome assembly by testing the presence of a set of single-copy orthologs conserved across a wide range of taxa. However, BUSCO is slow particularly for large genome assemblies. It is cumbersome to apply BUSCO to a large number of assemblies. Results Here, we present compleasm, an efficient tool for assessing the completeness of genome assemblies. Compleasm utilizes the miniprot protein-to-genome aligner and the conserved orthologous genes from BUSCO. It is 14 times faster than BUSCO for human assemblies and reports a more accurate completeness of 99.6% than BUSCO’s 95.7%, which is in close agreement with the annotation completeness of 99.5% for T2T-CHM13. Availability and implementation https://github.com/huangnengCSU/compleasm. 1 INTRODUCTION With the recent advances in sequencing technologies, especially long-read sequencing, genome assembly has undergone a revolution in terms of quality and completeness. Long-read sequencing technologies have improved genome assembly by enabling the generation of more complete and complex genome assemblies, including the telomere-to-telomere assemblies and haplotype-resolved assemblies, which were previously challenging to assemble using short-read sequencing (Wenger et al. 2019, Nurk et al. 2022). With the development of new assembly algorithms and improvements in computational resources, the time required to assemble a human genome has decreased significantly. It is now possible to assemble a human genome within 8 h (Cheng et al. 2021, 2022). Evaluating assembly completeness is a critical process that provides valuable information about the accuracy and reliability of the assembly and allows for finding potential errors such as missing or misassembled regions. Various tools have been developed for evaluating the completeness of genome assemblies. QUAST (Gurevich et al. 2013) is a widely used assembly quality evaluation tool by mapping the contigs/scaffolds of the assemblies to the reference genome and then reports the number of misassemblies, gaps, and various contiguity metrics including N50 and L50, but it is not applicable to new species. Benchmarking Universal Single-Copy Orthologs (BUSCO) is a popular tool for evaluating the completeness of assemblies (Simão et al. 2015, Manni et al. 2021). BUSCO utilizes a set of conserved single-copy orthologs that are expected to be present in the assemblies and employs gene predictors to confirm the presence of these genes in the assembly to assess the assembly completeness. However, BUSCO may underestimate completeness. For example, for the telomere-to-telomere (T2T) CHM13 assembly (Nurk et al. 2022), BUSCO reports a completeness of only 95.7%, but applying BUSCO to annotated protein-coding genes gives a completeness of 99.5%. In addition, BUSCO is inefficient. Assessing the completeness of a human genome assembly using BUSCO can take around 7 h, which is close to the time required for assembling a human genome. With the increasing complexity and size of genome assemblies, there is an urgent need for more efficient and accurate completeness evaluation tools. In this study, we will introduce compleasm, pronounced/kəmˈplēzəm/, an efficient tool for assessing assembly completeness. It reimplements some logic behind BUSCO but replaces the core protein-to-genome alignment algorithm with miniprot (Li 2023). The evaluation of real datasets demonstrates that compleasm has a significant speed-up compared to BUSCO while achieving higher accuracy. 2 MATERIALS AND METHODS Compleasm first downloads the corresponding BUSCO lineage dataset from https://busco-data.ezlab.org/v5/data (Manni et al. 2021, Zdobnov et al. 2021) to evaluate the input genome or assembly. Each lineage dataset includes hundreds to over ten thousand near-universally distributed single-copy gene groups with each group consisting of multiple protein sequences from different species (Simão et al. 2015). To evaluate the completeness of an input assembly, we map the protein sequences of these single-copy gene groups to the assembly and quantify the occurrences of these single-copy genes in the assembly. In cases the assembly is incomplete, certain single-copy genes cannot be aligned. On the contrary, if the assembly contains falsely duplicated genome regions, some single-copy genes will be aligned to multiple positions in the assembly. Miniprot is a fast aligner for mapping protein sequences to the genome. Compared to similar tools, miniprot has the advantages of shorter runtime and accurate detection of splice junctions and frameshifts. Compleasm performs one round of miniprot, while BUSCO performs two rounds of MetaEuk (Levy Karin et al. 2020) with different parameters for high enough sensitivity. Due to the similarity of sequences in the genome, even if some regions are missing in the assembly, the protein sequences of the corresponding single-copy genes may be aligned to other paralogs in the assembly at lower identity. This may overestimate the completeness. Like BUSCO, we employ HMMER3 (Mistry et al. 2013) to confirm orthology. Only matches above the score cutoff defined in lineage files will be retained. Since each single-copy gene may have multiple protein sequences, we choose the protein sequence with the highest hmmersearch score to represent the single-copy gene group. With paralogous gene matches filtered out, the remaining matches are categorized into one of four types: complete and single-copy, complete and duplicated, fragmented, or missing. A gene is considered missing if it has no alignment after HMMER filtration. It is fragmented if all its alignments are shorter than a length threshold defined by BUSCO. The rest of genes are considered complete. A complete gene is considered to have a single-copy in the assembly if it only has one alignment, or duplicated if it has multiple alignments. Compleasm reports the proportion of genes falling into each of the four categories as the assessment of assembly completeness. 3 RESULTS To evaluate the performance of compleasm, we compared it with BUSCO on various species, including seven reference genomes of model organisms: Homo sapiens, Mus musculus, Arabidopsis thaliana, Zea mays, Drosophila melanogaster, Caenorhabditis elegans, and Saccharomyces cerevisiae. Furthermore, we evaluated compleasm and BUSCO on 81 PacBio HiFi assemblies of Metazoa and 22 PacBio HiFi assemblies of Viridiplantae obtained from the Darwin Tree of Life Project. To analyze the frameshifts reported by compleasm, we ran compleasm on several Nanopore assemblies and PacBio assemblies. The details of the datasets can be found in Supplementary Table S1. For BUSCO, we used the option “-m genome” and the default gene predictor MetaEuk. 3.1 Evaluation of model organism reference genomes Table 1 shows the comparison of compleasm and BUSCO on the reference genomes of seven model organisms. For A.thaliana and S.cerevisiae reference, the results of compleasm and BUSCO shows a difference of completeness (single-copy and duplicated) <1%. For D.melanogaster, C.elegans, and M.musculus, the differences between the assessment results of compleasm and BUSCO range from 1% to 3%. When evaluating the reference genomes of H.sapiens and Z.mays, compleasm and BUSCO show a large difference exceeding 3%. For most reference genomes, compleasm reports higher completeness than BUSCO. [Table: Comparison of compleasm and BUSCO on the reference genomes of seven model organisms.] In the assessment of the H.sapiens reference genome (T2T-CHM13), compleasm reported completeness of 99.6% compared to 95.7% reported by BUSCO. Among them, 562 complete genes were only reported by compleasm. We subsequently evaluated these compleasm-specific