Licensing¶

BacFlux is MIT licensed. Some of the tools, model sets and databases it can invoke are not, and this page is the one place their terms are written down.

It records what each licensor publishes, and draws no conclusion about whether a particular use is permitted — that depends on who you are and what you are doing, and it is yours to establish with the licensor.

Terms used on this page


HMM	hidden Markov model, a statistical profile used to recognise a protein family
ICE	integrative and conjugative element — moves itself between cells
IME	integrative mobilisable element — needs a helper element to move

BacFlux's own code¶

The repository contains no third-party database, model set or sequence file. It ships code and URLs. Every database a run needs is either one you already hold and point the config at, or one the workflow fetches for you, on your machine, from the publisher — the arrangement bakta_db, gtdbtk_db, platon_db and the CARD link have always had, and the one every optional mobilome layer uses too. See Databases for what that means in practice.

Data and code are two different problems¶

	A database or model set	Source code
What BacFlux distributes	a URL in `config/config.yaml`	the file itself, in the repository
Who obtains it	you, under your own agreement with the licensor	anyone who clones the repository
Does a config switch settle it?	yes — nothing is shipped in the first place	no

Why default-off settles one and not the other

A config switch controls what runs. It does not control what is distributed. Share-alike terms — the "if you pass it on, pass it on under the same terms" clause in CC BY-NC-SA — attach to source that ships in the repository, so a CC BY-NC-SA file would carry its terms into BacFlux whether or not any switch ever ran it. The only two remedies are to not copy it, or to relicense the whole workflow. This project took the first.

Running such a tool is a different act from redistributing its source, and does not affect BacFlux's own licence.

The default path¶

VirSorter2 and not geNomad is the default virus caller, and the reason is the licence rather than the science (workflow/rules/shared/00_common.smk:104-111). The licence types a default run depends on:

Component	Licence type
VirSorter2 — default virus caller	GPLv2
Platon — plasmid caller, always on	GPLv3
CheckV — grades whatever the virus caller found	LBNL BSD

Those three are the components this repository states a licence type for. Every other tool BacFlux invokes — on the default path or inside the mobilome module — is distributed under its own licence by its own authors; CITATIONS.md lists them all, and each tool's repository carries the licence itself. The databases every run needs — Bakta, GTDB, CARD and the rest — are fetched the same way, under their publishers' terms; Databases says what to download and from where.

Components you switch on yourself¶

Each of these stays off until you set the key in the middle column.

Component	Turned on by	Licence as published	Read from
geNomad — alternative virus caller, and the second opinion in the plasmid stage	`phage.caller: genomad`	Berkeley Lab licence, academic / non-commercial use only — "User must be an accredited academic institution"; commercial use requires a separate LBNL licence	the raw `LICENSE` file, checked 2026-07-22
CONJScan models — the HMM profiles and system definitions that detect conjugation machinery	`mobilome.run: true` (fetched whenever the module runs)	CC BY-NC-SA 4.0, Institut Pasteur / CNRS	the package's own `metadata.yml`
MacSyFinder — the engine those models run in	same	GPLv3	—
ICEscan models — the optional second model set, a fork of CONJScan distributed inside ICEfinder2's bundle	`mobilome.icescan.run: true`	CC BY-NC-SA 4.0, Institut Pasteur / CNRS — the same terms as the models it forks	the package's own `LICENSE` file
TnCentral — names transposons and integrons	`mobilome.tncentral.url` or `.dir`	an "All Rights Reserved" notice, and no terms page at all	the publisher's site, checked 2026-07-28
ICEberg 3.0 — names the ICE and IME candidates	`mobilome.iceberg.urls` or `.dir`	no licence, terms or reuse statement anywhere; every page carries only "Copyright © 2023 All Rights Reserved by Microbial Bioinformatics Group in MML, SJTU."	the publisher's site, checked 2026-07-28
ISOSDB — counts the IS copies the assembly collapsed	`mobilome.isosdb.fasta_url` + `family_map_url`, or `.dir`	MIT, in the pseudoR repository	the repository

MacSyFinder is not CONJScan: the engine is GPLv3, the models it loads are CC BY-NC-SA 4.0, and it is the models that BacFlux fetches. geNomad's bioconda recipe is tagged BSD-4-Clause, which does not match the raw LICENSE file — the file is what was read for the row above.

No published terms is not the same as permissive

TnCentral and ICEberg publish no terms of use. That is an absence of a statement, not a grant. Both layers are opt-in, neither is fetched unless you configure a source, and BacFlux redistributes neither.

ISfinder is not wired in¶

ISfinder is the reference catalogue of bacterial insertion sequences. Its terms are stricter than any other resource BacFlux touches:

"It is not permitted to download the ISfinder database without written authorization. Moreover, it is also not permitted at any time to distribute the database to third parties either individually or as part of web-based software."

— ISfinder terms of use, read 2026-08-17

So BacFlux never offers you a way to fetch it. Every other optional database in the mobilome module works the same way: the workflow ships a URL, you download under the publisher's terms. For ISfinder there is no URL anywhere in the configuration, because supplying one would route around an authorisation only you can obtain.

This does not cost you anything. ISEScan, which is how BacFlux actually finds insertion sequences, uses its own bundled profile HMMs and needs no ISfinder data.

Where it could still reach you. TnCentral serves several endpoints, and two of them bundle ISfinder content. Only nc/tn, the transposon set on its own, appears in the configuration; tn_in_is and prot/tn_is are named in the config comments solely to record that they are deliberately left out (config/config.yaml, the comment above mobilome.tncentral). Point mobilome.tncentral.url at one of those two and ISfinder's terms become yours to satisfy — which is a decision to make deliberately, not one to make by copying a URL out of a comment.

What a run tells you¶

The workflow reports this itself, not only in this document:

Selecting geNomad prints a notice at parse time, before any job starts (workflow/rules/shared/00_common.smk:123-127).
conjscan_models and icescan_models write the CC BY-NC-SA notice to the top of their own logs (workflow/rules/shared/80_mobilome.smk:596, :730).
Every opt-in download comes from an unversioned address, so each writes a PROVENANCE.txt beside the data. tncentral_db, iceberg_db and isosdb_db record the source, the fetch date, the checksum of the FASTA that actually arrived, the sequence count and the licence statement for that source; icescan_models records the package's own version line and profile count in place of a checksum. That file is the only record of which release you analysed — see Optional layers.

What this repository deliberately does not contain¶

No CC BY-NC-SA code. EBI's mobilome-annotation-pipeline is Apache-2.0, but three of its scripts — bin/ice_boundary_refinement.py, bin/map_tools/icefinder_process.py and bin/prescan_to_fasta.py — derive from ICEfinder2 and carry CC BY-NC-SA headers. None of it is copied here. What was taken from that pipeline is decisions: that ICEscan exists and is worth running, the Sequence Ontology terms for the mobilome GFF, the contig_id|mge_type-start:end element ID format, the discard-with-reason file pattern, and the rule that a gene counts as inside an element only when at least 0.9 of it lies within the boundaries. Those are facts and design choices, not expression. Reading source to understand an algorithm and then writing your own is not copying, and that is the line kept here.

Consuming that pipeline's mobilome.gff.gz output carries no licence consequence, which is why it is offered as an escape hatch for deeper mobilome work.

No vmatch. The att-site search computes maximal exact repeats in plain Python. bioconda's vmatch package declares license: Unknown / OTHER and vmatch.de is unreachable, so its terms cannot be established — which is weaker ground than a known-restrictive licence. That the substitute returns the same repeats was measured rather than assumed; see Validation.

Citing¶

Licensing and citation are separate obligations: a permissive licence still expects the citation. Citation covers how to cite BacFlux and the tools that do the work.

Standing caveat

This page records what each licensor published, as read on the dates given. Licences change and pages move. Before relying on any of it, read the current terms at the source. Nothing here is legal advice.