Mobilome module¶
Stage 08.mobilome answers one question, once for every antimicrobial resistance
gene BacFlux finds:
Is this gene embedded in a mobile genetic element, and if so, how transferable is that element?
It is optional and off by default; everything below assumes you have switched it on (Turning it on).
Four terms this section uses constantly
An ICE (integrative and conjugative element) sits in the chromosome and carries its own conjugation machinery, so it can move itself into another cell. An IME (integrative mobilisable element) sits in the chromosome and carries a relaxase — the enzyme that nicks the DNA to start a transfer — but no apparatus of its own, so it needs a helper element to move. An att site is the short direct repeat left at each end of an element where it integrated; finding the pair is how an element's true edges are established.
Three more abbreviations appear throughout: IS is an insertion sequence, the smallest mobile element, carrying only the genes it needs to move itself; CDS is a coding sequence, the stretch of DNA that codes for one protein; and EFSA is the European Food Safety Authority.
Why a gene's address matters as much as its name¶
An AMR gene list tells you what an isolate can resist. It does not tell you whether that resistance can move. A chromosomal efflux pump that every member of the species carries, and a blaCTX-M sitting on a conjugative plasmid, produce the same kind of line in an ABRicate table and mean very different things — only the second can end up in another bacterium. That is the distinction regulators ask about: intrinsic resistance versus acquired.
The module produces supporting evidence for that judgement, never the judgement itself. The mobility ladder sets out the framing, the EFSA definitions behind it, and why tier 1 says candidate.
The mobility ladder¶
Every assessed gene gets exactly one tier. The label is written into the output table next to the number, so nobody reading the TSV has to look a tier up.
| Tier | mobility_tier_label |
Context |
|---|---|---|
| 1 | intrinsic_candidate |
chromosomal, nothing mobile nearby |
| 2 | expression_modulation_not_mobilisation |
an IS upstream, on the gene's strand — expression, not mobilisation |
| 3 | composite_mobilisable_within_cell |
between two IS copies of one family: a composite transposon |
| 4 | named_element_mobilisable |
inside a curated unit transposon or integron cassette |
| 5 | mobilisable_needs_helper |
on a plasmid, or inside an IME |
| 6 | predicted_self_transmissible |
inside an ICE, or on a conjugative plasmid |
The rungs are tested top down and the first match wins, so a curated transposon on a plasmid scores 5 or 6 rather than 4. That order, the three distance windows behind it, and how far to trust each rung — tier 4 is measured on two genomes, tier 5's IME route on none — are all on The mobility ladder.
What runs when you turn it on¶
The module adds 22 Snakemake rules, of which 10 run on the default configuration. The other 12 belong to four optional layers, each of which stays off until you give it a source (Optional layers).
flowchart LR
C([contigs_final.fasta<br/>02.assembly])
FAA([bakta .faa and .gff3<br/>04.annotation])
TAX([GTDB-Tk call<br/>03.taxonomy])
PLA([Platon calls<br/>06.plasmids])
CL[contig_lengths]
IS[isescan<br/>insertion sequences]
IST[isescan_table]
AFP[amrfinderplus]
ORG[amrfinder_organism<br/>unlocks point mutations]
CJ[conjscan<br/>relaxase, coupling protein,<br/>mating apparatus]
CJI[conjscan_ice<br/>ICE and IME calls]
REP[mobilome_replicons]
C --> CL
C --> IS --> IST
FAA --> AFP
FAA --> CJ --> CJI
TAX --> ORG --> AFP
PLA --> REP
CO[amr_mge_colocalisation]
CL --> CO
IST --> CO
AFP --> CO
CJI --> CO
REP --> CO
CO --> O1([sample_amr_mobility.tsv])
CO --> O2([sample_amr_mobility_audit.tsv])Three things follow from that picture:
- Two new tools, both from bioconda: ISEScan v1.7.3 for insertion sequences, and MacSyFinder v2.1.6 running the CONJscan models for conjugation machinery. Both are given up to 8 threads per sample.
- No new database path to configure. AMRFinderPlus is already installed by Bakta
and reads its database from
{bakta_db}/amrfinderplus-db/latest; the CONJscan models are fetched at run time, so the machine needs network access on the first run. - Stage 08 consumes earlier stages rather than repeating them. It reads the
assembly, the Bakta proteins and GFF3, the GTDB-Tk species call, and Platon's
replicon calls from
06.plasmids.
The ICE/IME caller, the composite-transposon test and the att-site search run in
all four modes, including illumina. The one leg that is mode-restricted is the
optional ISOSDB copy-number layer, which maps reads and therefore needs them —
illumina and hybrid only.
The deliverable is one row per AMR gene in {sample}_amr_mobility.tsv, with an audit
file beside it giving a reason for every gene left without context, every candidate
structure rejected and every confidence cap applied. Column by column:
Reading the output.
Why the mobilome spans three stages¶
Plasmids and prophages are mobile genetic elements too, so the output layout can read
as if 08.mobilome/ were "the mobilome" and 06.plasmids/ and 07.phages/ were
something else. They are not. The split is by what question is being asked, not by
whether the element is mobile.
| Stage | Question | Runs |
|---|---|---|
06.plasmids, 07.phages |
what replicons and prophages are in this genome? (detection) | always |
08.mobilome |
is each AMR gene in a mobile element, and how transferable? (interpretation), plus IS, transposon, integron and ICE detection | opt-in |
Renaming 08.mobilome was considered and deliberately rejected in favour of
explaining it. See Plasmids and
Prophages for the detection stages.
Why it is off by default¶
Scope. Most characterisation rounds want the AMR gene list, not the mobility argument behind it. The always-on AMR legs — ABRicate on the contigs, and the CARD read screen in the modes that have short reads — are unaffected by this switch (Antimicrobial resistance).
Cost. Ten extra rules on the default path, three of them per-sample tool runs at up to 8 threads, plus a one-off model download.
Before you switch it on. Turning the module on fetches the CONJscan model
package, which Institut Pasteur / CNRS license under CC BY-NC-SA 4.0. BacFlux
never ships it — you download it, under your own agreement with the licensor, exactly
as with bakta_db and the other user-supplied databases. Licence types for every
tool, model set and database are listed on Licensing.
Three things to know before reading the output¶
1. On a fragmented assembly a located IS count is a lower bound. IS
elements are a leading cause of contig breaks: identical copies collapse in the
assembly graph, so the assembler cannot tell which flank belongs to which copy and
breaks there. The AMR gene and its flanking IS regularly land on different contigs.
Every row therefore carries dist_to_contig_end, is_at_contig_boundary and
spans_contigs, and the IS summary reports what fraction
of IS calls sit within 100 bp of a contig end (mobilome.contig_boundary_bp). Read
those before reading the counts.
2. Never write "transmissible". Tier 6 is predicted_self_transmissible, and the
prediction rests on finding conjugation machinery, not on watching a transfer happen.
3. Confidence is tiered, never a bare call. IS calling is imperfect even on
manually curated genomes: benchmarking four tools against a curated E. coli
annotation, Puterová & Martínek (2021) put the "improbable or not an IS element"
discovery rate between 0% and 23.7% depending on the tool, with ISEScan — the caller
BacFlux uses — at 8.0% (2.2% on their larger ISbrowser set). So every row carries
confidence (high / medium / low), and anything resting on a contig boundary or on a
partial hit is capped at low.
The module was validated on closed genomes and then measured again on deliberately fragmented ones — 40 benchmark genomes cut to ~150 kb, ~50 kb and ~20 kb N50, re-annotated and re-run end to end:
| closed | 150 kb N50 | 50 kb N50 | 20 kb N50 | |
|---|---|---|---|---|
| curated ICEs detected (of 18) | 15 | 15 | 15 | 12 |
| median called ÷ true element length | 0.55 | 0.45 | 0.34 | 0.22 |
ICE/IME calls at confidence: high |
24% | 19% | 10% | 3% |
The class survives fragmentation; the extent does not. A low-confidence call on a draft means the assembly could not support a stronger claim, not that the call is wrong. Which columns to read and what not to conclude: Draft assemblies.
Where to go next¶
| Page | What it covers |
|---|---|
| The mobility ladder | the six tiers in full, the IS-inside-CDS case, and the intrinsic-versus-acquired framing with its references |
| Turning it on | the one switch, and what each of the four optional layers needs before it will run |
| Optional layers | what each layer adds, what it costs, and where its data comes from |
| Reading the output | every column of {sample}_amr_mobility.tsv, the audit trail, and the rest of 08.mobilome/ |
| Draft assemblies | what fragmentation does to a call, and which columns still hold |
| Tuning | symptom to knob |
| Worked example | one genome end to end, with the real rows |
| Validation | the ICE and IME pilots, the negative control, the head-to-head against the EBI pipeline, and what the benchmark does not show |