Sign in
BioinformaticsNucleic Acids Research
Peer-reviewed journal article

PanGBank: a large-scale resource of precomputed microbial pangenomes built with PPanGGOLiN

Jean MainguyiD, Téo LemaneiD, Adelme BaziniD, Jérôme ArnouxiD, Guillaume GautreauiD, Claudine MédigueiD, Alexandra CalteauiD, David VallenetiD

Abstract

PanGBank (https://pangbank.genoscope.cns.fr) is a comprehensive open-access database providing precomputed prokaryotic pangenomes at a broad taxonomic scale. Built upon PPanGGOLiN partitioned pangenome graphs, PanGBank addresses the growing need for large-scale comparative genomics through a standardized, regularly updated, and fully accessible resource. The initial release comprises two complementary collections covering more than 4,600 prokaryotic species from the Genome Taxonomy Database (GTDB), encompassing over 393,000 genomes: GTDB all, maximizing taxonomic and environmental diversity through the inclusion of MAGs and SAGs, and GTDB refseq, focusing on high-quality, annotation-rich genomes. Each species-level pangenome integrates graph-based statistical partitions into persistent, shell, and cloud gene families, together with regions of genomic plasticity (panRGP) and co-localized functional modules (panModule). PanGBank offers multiple access modes, including a REST API, a command-line interface (PanGBank-cli), and an interactive web interface. By combining large-scale pangenome resources with advanced graph-based analyses, PanGBank provides a scalable framework for exploring microbial diversity, genome evolution, functional variation, and the dissemination of adaptive traits across prokaryotic populations, as illustrated by a use case on Acinetobacter baumannii pangenome investigating the distribution and evolution of antimicrobial resistance determinants. Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=63 SRC="FIGDIR/small/742796v1_ufig1.gif" ALT="Figure 1"> View larger version (23K): org.highwire.dtl.DTLVardef@12c441org.highwire.dtl.DTLVardef@12b079org.highwire.dtl.DTLVardef@ffd7f8org.highwire.dtl.DTLVardef@bc013f_HPS_FORMAT_FIGEXP M_FIG C_FIG

02 /Specification & Classification

Click any tag to browse matching papers

In-depth summary

✦ Multi-agent verified full-text synthesis

Claude Opus 5.5 · xhigh effort · 8 Nov 2026 · from the full text of version 1 (9 Aug 2026) · CC BY 4.0

Precomputed PPanGGOLiN pangenomes for over 4,600 prokaryotic species (Genome Taxonomy Database clusters), with plasticity regions and modules, downloadable via web, API and command line for projection onto new genomes.

Genome databases mostly represent a species by single assemblies, and the paper notes that precomputed pangenome resources are scarce: panX and ProPan have not been updated since release and offer no downloads, and EDGAR and MBGD use only curated complete or near-complete genomes. PanGBank adds downloadable partitioned pangenome graphs with plasticity and module predictions for thousands of GTDB species, including metagenome-assembled and single-cell genomes, with updates planned for each GTDB release.

Key results

  1. ShownThe first release is built from GTDB release 232 and has two collections. GTDB all (GenBank plus RefSeq) covers over 4,600 prokaryotic species. GTDB refseq (RefSeq only, original annotations kept) covers 2,044 species.

    In both collections, most pangenomes contain fewer than 50 genomes. Only about one hundred species have more than 1,000 genomes. 24 species (GTDB refseq) and 27 (GTDB all) went over the 2,000-genome cap and were dereplicated; these are mainly bacteria of public health importance. GTDB all has more regions of genomic plasticity (RGPs), modules and gene families (Table 1). · PanGBank content; Genome dereplication

  2. ShownIncluding metagenome-assembled genomes (MAGs) and single-cell amplified genomes (SAGs) gives broader taxonomic coverage. GTDB all spans 55 phyla, and around 40% of those phyla are covered only by environmental genomes. GTDB refseq is almost entirely isolate genomes.

    About 40% of GTDB all genomes are MAGs or SAGs. GTDB all holds about 150 archaeal pangenomes, which the authors attribute largely to the environmental genomes. Bacterial pangenomes make up most of both collections (Fig. 3A to 3C). · PanGBank content

  3. ShownAfter completeness filtering, the GTDB all collection (about 40% MAGs or SAGs) partitions much like the isolate-based GTDB refseq collection. The proportions of persistent, shell and cloud genomes agree strongly between the two collections.

    Some GTDB all pangenomes have slightly lower persistent fractions; the median decrease across matched species is 0.4 percentage points (Supplementary Figs S1, S2). Filters: completeness is the higher of the CheckM v1 and v2 values; genomes below 70% are dropped; species are dropped if their GTDB representative is below 85%; each species needs at least 15 genomes. · PanGBank content; Input filtering and species selection

  4. ShownIn the Acinetobacter baumannii pangenome, antimicrobial resistance (AMR) gene families sit mostly in the variable genome. Three insertion hotspots found in more than 500 genomes (spots 7, 11 and 47) are rich in AMR-carrying RGPs. Spot 47, at the comM insertion site, holds the most diverse AMR repertoire.

    AMRFinderPlus v4.2.7 found 123 AMR-associated gene families across 2,002 genomes: 79 cloud, 39 shell and 5 persistent. Half concern beta-lactam or aminoglycoside resistance. Spots 7, 11 and 47 occur in over 500 genomes, and more than 40% of their RGPs carry AMR genes. Spot 47 holds 33 AMR gene families, including those of AbaR1 from strain AYE. · PanGBank use case

  5. SuggestedIn A. baumannii, PPanGGOLiN's RGP clustering and projection recover known resistance-island structure. Spot 47 splits into 56 RGP clusters. In dog isolate ABO21-A003, the RGP with the most AMR genes maps to spot 31, the AbGRI2 island.

    At max grr 0.8, spot 47 gives 56 RGP clusters, close to the 53 genetic configurations described by Bi et al. ABO21-A003 is not in the pangenome; after projection it had four RGPs with AMR genes. Three fall in spots 7, 11 and 47. The fourth, CP136181.1 RGP 5, falls in spot 31 and matches AbGRI2 (reported by André et al.), with three blaTEM1 and two aphA1 copies. · PanGBank use case

Methods, data and reproducibility

A Nextflow workflow (PanGBank-wf) filters genomes on CheckM completeness, merges GTDB species split under one base name using Skani average nucleotide identity and alignment fraction, dereplicates species above 2,000 genomes via a Mash/Quicktree tree, then runs PPanGGOLiN with panRGP and panModule per species; results are served by a web site, REST API, Python SDK and command-line interface (CLI).

  • GTDB all collection: GTDB release 232 genomes from NCBI GenBank and RefSeq, downloaded as FASTA, including MAGs and SAGs · Bacteria and Archaea (about 150 archaeal pangenomes) · n = more than 4,600 species
  • GTDB refseq collection: GTDB release 232 RefSeq genomes downloaded in GenBank format with original annotations, almost all isolates · prokaryotes (dominated by bacterial pangenomes) · n = 2,044 species
  • Acinetobacter baumannii pangenome from the GTDB refseq collection v2.0.0, screened for AMR genes · Acinetobacter baumannii · n = 2,002 genomes
  • Genome of A. baumannii strain ABO21-A003, isolated from a dog in a veterinary intensive care unit and absent from the PanGBank pangenome · Acinetobacter baumannii · n = 1 strain · CP136181.1 (sequence ID printed in an RGP name)
Methods & tools
PPanGGOLiN, panRGP, panModule, Nextflow (nf-core guidelines), CheckM v1 and CheckM v2, Skani, Mash, Quicktree, ncbi-datasets-cli, MultiQC, AMRFinderPlus v4.2.7, CGView
Software & environment
PanGBank-wf v0.1.4, PanGBank-api v0.5.0 and PanGBank-cli v0.2.0 archived in Software Heritage; a MultiQC report on the site lists software versions; CLI on PyPI and Bioconda; use-case notebook runs via Docker, Conda or Google Colaboratory; AMRFinderPlus v4.2.7 (database 2026-05-15.1).
Reproducibility gaps
The Software Heritage identifiers (SWHIDs) are blank in the supplied text; Table 1 values and the supplementary figures and table are not included; no compute or runtime figures are reported.
Limitations (4)
  • GTDB assigns each genome to a species through a representative genome, so close genomes can land in different species. Merging only covers species that share a base name with a letter suffix, because comparing all pairs of GTDB species by nucleotide identity is intractable. Other split species stay uncorrected. · stated by the authors
  • The authors flag low-quality MAGs and SAGs as a possible source of bias in pangenome reconstruction. They reduce it with completeness filters: genomes under 70% are removed, and species are dropped if their GTDB representative is under 85%. · stated by the authors
  • Evidence that the resource is useful comes from one species (A. baumannii) and one external genome. The agreement with prior work is a comparison of counts (56 RGP clusters vs 53 configurations from Bi et al.). There is no systematic benchmark of partitions or RGP calls across the collections. · noted in this summary
  • Gene families are not yet systematically annotated for function across collections; the authors list this as future work with PANORAMA. The AMR use case needed a separate AMRFinderPlus screen. GTDB all genomes were fetched as FASTA, so they lack the original NCBI annotations that GTDB refseq keeps. · noted in this summary

Who should read:Read this if you study prokaryotic gene content variation and want downloadable PPanGGOLiN pangenomes, plasticity hotspots, or a reference pangenome to project a newly sequenced genome onto.

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Similarity Map
SelectedSame fieldCross-field
THISBoying '24Arnoux '26Zhou '18Le '24Beier '22Noll '23
76% semantic matchGenomics
Sun Boying, Liubov Pashkova, Pascal A. Pieters+6 more·Nucleic Acids Research·Oct 19, 2024· 14 citations