BUSCO

BUSCO
Name	BUSCO
Developer	Fernando de la Cruz?
Released	2015
Latest release version	5.x
Programming language	Python, C, Bash
Operating system	Linux, macOS
License	MIT-style

BUSCO BUSCO is a bioinformatics tool designed to assess the completeness of genome assemblies, annotated gene sets, and transcriptomes using evolutionarily informed expectations about conserved single-copy orthologs. It provides per-sample estimations that help researchers in genomics, comparative genomics, and phylogenomics determine whether data sets include the expected repertoire of conserved genes. BUSCO is widely used in pipelines for de novo assembly evaluation, annotation validation, and large-scale biodiversity projects.

Overview

BUSCO evaluates biological sequence data against curated sets of orthologous groups derived from comparative analyses of model organisms and major taxa. The approach relies on lineage-specific databases constructed from well-studied species such as Homo sapiens, Mus musculus, Drosophila melanogaster, Caenorhabditis elegans, Arabidopsis thaliana, Saccharomyces cerevisiae, Escherichia coli, Zea mays, Oryza sativa, Danio rerio and others to provide benchmarks for completeness. Results are reported as proportions of Complete, Duplicated, Fragmented, and Missing orthologs, offering interpretable metrics for projects ranging from single-genome efforts to consortia-level initiatives like The Human Genome Project and Earth BioGenome Project. Developers and users often integrate BUSCO outputs alongside measures from tools such as QUAST, Canu, SPAdes, MAKER (bioinformatics), and Augustus.

Methodology

BUSCO’s core methodology scans input nucleotide or protein sequences for matches to curated single-copy ortholog groups, leveraging hidden Markov models (HMMs) and sequence aligners. Lineage datasets are built from orthology inference frameworks and resources like OrthoDB, OrthoMCL, EggNOG, and databases constructed from genomes of species such as Saccharomyces cerevisiae, Arabidopsis thaliana, Drosophila melanogaster, Anolis carolinensis, Gallus gallus, Xenopus tropicalis, Mus musculus, Bos taurus, Canis lupus familiaris, Equus caballus, Sus scrofa, Pan troglodytes, Gorilla gorilla, Taeniopygia guttata, and Strongylocentrotus purpuratus. For nucleotide inputs, gene prediction components from tools like AUGUSTUS are invoked to produce predicted proteins, while HMMER family software performs HMM searches. Matches are classified using criteria informed by the Benchmarking Universal Single-Copy Orthologs concept and ortholog presence patterns across taxa. BUSCO distinguishes true single-copy ortholog detection from paralogy and assembly artifacts by integrating taxon-aware thresholds and sequence-quality heuristics.

Applications

BUSCO is applied across genomics workflows including assembly quality control, annotation pipeline validation, transcriptome completeness checks, and metagenomic bin assessment. Researchers in projects associated with 1000 Genomes Project, Genome 10K, i5K, Bird 10,000 Genomes Project, Marine Microbial Eukaryote Transcriptome Sequencing Project and biodiversity surveys employ BUSCO to compare assemblies from assemblers like Canu, Flye, SPAdes, MEGAHIT, and SOAPdenovo2. Conservation genomics teams working on species such as Gorilla beringei, Pan paniscus, Puma concolor, and Elephas maximus use BUSCO to validate draft genomes before downstream analyses with pipelines invoking MAKER (bioinformatics), BRAKER, PASA, and EVM (software). BUSCO outputs also inform phylogenomic matrix assembly for studies using methods from RAxML-NG, IQ-TREE, MrBayes, and ASTRAL.

Performance and Benchmarking

Benchmarking studies compare BUSCO’s sensitivity and specificity to other completeness metrics across simulated and empirical data sets. Comparisons often involve tools and resources such as QUAST, CEGMA, CheckM, MetaQUAST, REAPR, and ALE (software). BUSCO generally provides robust, lineage-aware estimates that correlate with assembly contiguity metrics like N50 and L50 from assemblers such as Canu and SPAdes, but its performance depends on the appropriateness of the chosen lineage dataset. Large taxonomic breadth datasets increase sensitivity for deeply conserved genes but may reduce resolution for recent radiations; conversely, narrow lineage datasets enhance detection in clade-specific contexts. Runtime and memory depend on input size, HMMER search complexity, and gene-prediction steps; parallelized deployments on compute clusters with resource managers like SLURM or PBS (software) are common practice.

Software Implementations and Tools

BUSCO is implemented in Python with dependencies on HMMER, BLAST or DIAMOND, and gene prediction packages such as AUGUSTUS and MetaEuk for eukaryotic inputs. Workflow managers and integrators often wrap BUSCO in reproducible pipelines using Snakemake, Nextflow, CWL, and container technologies like Docker and Singularity for portability across systems employed by institutions such as European Bioinformatics Institute, National Center for Biotechnology Information, Wellcome Sanger Institute, Broad Institute, and JGI (Joint Genome Institute). Community extensions and analyses interoperate with visualization tools like MultiQC, Krona, and plotting libraries used in projects from Galaxy (web platform) instances to custom R and Python notebooks.

Limitations and Criticisms

Critics note several limitations: reliance on curated lineage datasets can bias completeness estimates when target taxa are poorly represented among reference genomes (issues observed in understudied clades such as some Protists, Nematoda lineages, or deep-branching Fungi); single-copy ortholog assumptions can be violated by lineage-specific duplications in groups such as Teleostei or Angiospermae; and gene-prediction inaccuracies can inflate Fragmented or Missing categories, particularly for highly divergent gene models in taxa like Platyhelminthes or Myxozoa. BUSCO scores should therefore be interpreted alongside complementary metrics from tools like QUAST and manual curation in annotation platforms developed at institutions such as Ensembl or RefSeq to avoid overconfidence in draft assemblies.

Category:Bioinformatics software