ESKAPE pathogen prophage repertoire : PHORAGER output collection

This archive contains the key prophage-mining outputs generated by PHORAGER (Prophage Hunting, vOTU Retrieval, Annotation and Genomic ExploRation) applied to a large collection of publicly available ESKAPE pathogen genomes.

- Pipeline: PHORAGER v0.5.0-beta https://github.com/aponsero/PHORAGER
- Associated manuscript: Defining the ESKAPE pathogen prophage repertoire with PHORAGER (Dyball et al. bioRXiv)
- Generated by: the PHORAGER `bacterial`, `prophage`, `annotation`, and `summarize` workflows, run with default parameters.

------------------------------------------------------------------------------------------------------------------------------------------------

1. What is in this archive

The collection is split into one ZIP file per ESKAPE pathogen. Each ZIP unpacks to a single folder containing five files for that species:

Species included (six ZIPs):

- Enterococcus faecium
- Staphylococcus aureus
- Klebsiella pneumoniae
- Acinetobacter baumannii
- Pseudomonas aeruginosa
- Enterobacter spp.

Folder / file naming. Folders follow `<Species>_<DD.MM.YY>`, where the date is the NCBI download/build date for that species' genome set (these differ between species, e.g. most ESKAPE genomes were retrieved 10 Oct 2025, K. pneumoniae on 22 Oct 2025). Files within a folder are prefixed with the same `<Species>` token.

------------------------------------------------------------------------------------------------------------------------------------------------

2. File descriptions

2.1 `<Species>_genomes_metadata.tsv` : NCBI input metadata

Tab-separated metadata for the bacterial genomes downloaded from NCBI and used as input to PHORAGER. This file is produced by the NCBI download step, not by PHORAGER.

2.2 `<Species>_bacterial_genome.tsv` : per-genome summary

One row per genome that passed CheckM2 quality assessment. Columns (exact, as emitted
by the pipeline):

> Note: `Prophage_count` is derived from the filtered (non-dereplicated) prophage set,
> by attributing each prophage to the genome named before the first `_ctg` in its ID.

2.3 `<Species>_filtered_phage_set.fasta` : filtered prophage sequences

Nucleotide FASTA of all prophage sequences that passed PHORAGER's annotation-based filtering, before vOTU dereplication. Each record corresponds to exactly one row in `prophage_table.tsv` (matched on the sequence ID). Sequence-ID format is described in §3.

2.4 `<Species>_final_representatives.fasta` : vOTU representatives

Nucleotide FASTA containing one representative sequence per vOTU cluster (95% ANI / 85% aligned fraction by default). These are extracted directly from `filtered_phage_set.fasta`, so every header here also appears in that file and equals a value in the `Cluster_representative` column of `prophage_table.tsv`. Use this file as the non-redundant, species-level prophage set for downstream analysis.

2.5 `<Species>_prophage_table.tsv` : per-prophage summary

One row per prophage in `filtered_phage_set.fasta`. Columns (exact, as emitted):

------------------------------------------------------------------------------------------------------------------------------------------------

3. Prophage sequence-naming convention

Input contigs are renamed by PHORAGER to `{genome_basename}_ctg{NNN}`, where `genome_basename` is the input genome's filename stem and `NNN` is a zero-padded contig number (3–5 digits, depending on contig count per genome). Prophage records then take one of two forms:

- Coordinate-based (merged/consolidated predictions):
  `{genome_basename}_ctg{NNN}_{start}_{end}`
  Coordinates are 1-based and inclusive, relative to the renamed contig.
  Example: `GCF_000123456.1_ctg001_4521_38790`
- Whole-contig predictions:
  `{genome_basename}_ctg{NNN}_complete`
  Used when a mining tool flagged the entire contig as prophage.

CheckV may append a trailing `_1` to trimmed proviral sequences; where present, that suffix is part of the FASTA header and the `Prophage_name` value.

------------------------------------------------------------------------------------------------------------------------------------------------

4. How the files link together

The four PHORAGER files share consistent identifiers, so they can be joined directly:

- Prophage → host genome: `prophage_table.Bacterial_host` = `bacterial_genome.Genome_name`.
- Prophage row → sequence: `prophage_table.Prophage_name` = FASTA header in `filtered_phage_set.fasta`.
- Prophage → its vOTU representative: `prophage_table.Cluster_representative`; the representatives themselves are in `final_representatives.fasta`.
- Genome → NCBI metadata: `bacterial_genome.Genome_name` = the accession column in `genomes_metadata.tsv` *(confirm which column — see §2.1)*.

------------------------------------------------------------------------------------------------------------------------------------------------

5. How these outputs were generated

Genomes were processed with PHORAGER v0.5.0-beta using default parameters. In summary
(see the associated manuscript for full methods and citations):

1. Bacterial QC & dereplication : CheckM2 quality filtering (default ≥90% completeness, <5% contamination) followed by dRep dereplication at the strain level (default 99.9% ANI).
2. Prophage mining & consolidation : geNomad and VIBRANT run in parallel; overlapping predictions consolidated into non-redundant coordinate ranges; sequences extracted.
3. Annotation & filtering : CheckV quality assessment; length/quality filtering; annotation with Pharokka (and Phold); structural-gene-based filtering. The `filtered_phage_set.fasta` is the output of this stage.
4. vOTU clustering : BLAST + CheckV ANI scripts, clustered at the species level (default 95% ANI / 85% aligned fraction). `final_representatives.fasta` holds the cluster representatives.
5. Summaries : `bacterial_genome.tsv` and `prophage_table.tsv`.

Tool versions: