rpoB (bacterial RNA polymerase subunit B)

rpoB encodes the beta subunit of bacterial RNA polymerase. Like gyrB it is a protein-coding marker, single-copy in almost all bacteria, and it assigns a far higher proportion of ASVs to species than a 16S V3–V4 amplicon does. Ogier et al. (2019) put it forward, together with the degenerate Univ_rpoB_deg primer pair, as a marker for describing bacterial community diversity by amplicon sequencing, and report 60–80% of rpoB OTUs assigned at species level against about 20% for 16S. It does not resolve strains: the reference stops at species, so neither does the output.

The MetaFlux path for rpoB is the ordinary amplicon path — PhiX removal, primer trimming, DADA2 denoising, length filtering, taxonomy. What is marker-specific lives in workflow/markers/rpoB.yaml: the reference is a set of full-length genes rather than pre-cut amplicons, it needs a one-time reformatting before DADA2 can read it, and only the RDP classifier is available.

At a glance

amplicon.type	rpoB (also accepted as rpob or RPOB — matched case-insensitively)
Reference database	FROGS RefSeq rpoB, build 20240707 (complete + chromosome NCBI RefSeq genomes)
Role of that database	both the taxonomy reference and the amplicon-length probe substrate
Reference content	full-length rpoB genes, roughly 4 kb, about 47,000 sequences
Probe mode	pcr — in-silico PCR recovers the amplicon window
Preparation	download + unpack, then a one-time header conversion to DADA2 format
Classifier	RDP only (amplicon.taxonomy.method: rdp)
Separate species step	none — the trainset itself carries species
Region extractor	none (no rpoB equivalent of Metaxa2 or ITSx)
Rank model	Domain;Phylum;Class;Order;Family;Genus;Species (no kingdom)
Primers	none are shipped; Ogier 2019 Univ_rpoB_deg F/R is the usual pair

Setting up a run

Set the marker type, point at the primer FASTAs, and choose the RDP classifier:

mode: amplicon

amplicon:
  type: rpoB
  primers:
    fwd: /path/to/Univ_rpoB_deg_F.fasta
    rev: /path/to/Univ_rpoB_deg_R.fasta
    orientation: fixed          # fixed | mixed
  expected_length: auto
  probe_length_stat:
    rpoB: p95                   # required for rpoB, auto or not
  extraction:
    enabled: false              # no rpoB region extractor
  taxonomy:
    method: rdp                 # rpoB has no SINTAX build
    min_boot: 80
    try_rc: true
    filter:
      enabled: true
      keep:    []
      discard: []               # no recommended default for rpoB

Primers are never supplied by MetaFlux: amplicon.primers.fwd and amplicon.primers.rev must both point at existing FASTA files, and the run stops at start-up if either is missing. Univ_rpoB_deg is degenerate, which is what makes it broad enough for community work. The same primer FASTAs are used both to trim the reads and to run the in-silico PCR described below, and both steps allow the mismatch rate set by amplicon.cutadapt.max_error_rate (default 0.2).

The shipped config template defaults to method: sintax

config/config.yaml ships with amplicon.taxonomy.method: sintax. That is fine for 16S/ITS/18S but not available for rpoB, so a run config copied from the template without changing that line stops at start-up, before any compute is spent:

[MetaFlux] marker rpoB has no SINTAX reference database, so
amplicon.taxonomy.method: sintax is not available for it. Use method: rdp.

The reference database

The reference is the FROGS RefSeq rpoB databank built on 2024-07-07 from complete and chromosome-level NCBI RefSeq genomes — the rpoB genes extracted from those genomes, about 47,000 sequences, each a full-length gene of roughly 4 kb.

Downloaded from	https://web-genobioinfo.toulouse.inrae.fr/frogs_databanks/assignation/rpoB/rpoB_bacteria_NCBI_refseq_genome_complete_and_chromosome_20240707.tar.gz
Fetch rule	fetch_rpob_archive → <out_dir>/logs/refdb/fetch_rpob_archive.log
Conversion rule	convert_rpob_to_dada2 → <out_dir>/logs/refdb/convert_rpob_to_dada2.log
Cached at	refdb/rpob/rpob_refseq_cc_20240707_domain-d_dada2.fa.gz

The cache root is references.refdb_root (default refdb). An entry under references is only needed to relocate a database that already exists elsewhere on disk — for rpoB that entry names the converted file:

references:
  rpob:
    frogs: /shared/dbs/rpob_refseq_cc_20240707_domain-d_dada2.fa.gz

The converted file is one MetaFlux builds itself, so a path given here is where MetaFlux writes it, and it rebuilds it there whenever the conversion script is newer than the file (see reference files MetaFlux builds itself).

Both rules run once, the first time an rpoB run needs the database, and are skipped afterwards because their output is already on disk. "domain-d" in the file name marks the file that writes the domain as d__ (below). The converted file of MetaFlux v2.4.0 and earlier, rpob_refseq_cc_20240707_dada2.fa.gz, wrote it as k__ and is no longer read, so the first rpoB run of a newer MetaFlux downloads the FROGS archive again (173 MB) and converts it once.

Why a conversion step exists

FROGS distributes the databank as a .tar.gz bundling three things: the FASTA, a BLAST index, and an RDP-classifier tree. fetch_rpob_archive downloads and unpacks it, keeps only the FASTA (the BLAST index files share the .fasta stem, so the rule picks the one true FASTA and skips the .properties file), and gzips it to a temporary refdb/rpob/rpob_frogs_raw.fasta.gz. Snakemake deletes that intermediate once the conversion has consumed it.

The FASTA is shaped for FROGS' own affiliation tool, not for DADA2. Every header is an accession followed by a lineage under a constant Root, in which each node carries a [id: N] database tag:

>WP_095092576.1 Root;k__Bacteria [id: 1];p__Bacillota [id: 2];...;s__Staphylococcus_simiae [id: 7]

DADA2's assignTaxonomy instead wants the header to be the lineage — semicolon separated, no accession, no per-node tags. convert_rpob_to_dada2 runs workflow/scripts/refdb/frogs_rpob_to_dada2.py, which makes exactly three edits per header and leaves sequence lines untouched:

1. drop the leading >ACCESSION Root;, so the lineage starts at k__Bacteria;
2. strip every [id: N] tag;
3. write the first rank as a domain: k__Bacteria becomes d__Bacteria.

The third edit is needed because FROGS writes the domain under the k__ prefix. This release carries NCBI's taxonomy as it was when FROGS built it (July 2024), before NCBI added kingdoms for prokaryotes, so every lineage has a domain and no kingdom. In MetaFlux's tables d__ is always the domain and k__ always the kingdom, so the rpoB pack names its first rank Domain, with the prefix d__, and the converted file says the same. The result is the file DADA2 trains on:

>d__Bacteria;p__Bacillota;c__Bacilli;o__Bacillales;f__Staphylococcaceae;g__Staphylococcus;s__Staphylococcus_simiae

The sequences are not changed; the conversion rewrites only the headers.

The converter refuses to run on an unexpected header shape

All three edits are anchored to the exact shape of the 2024-07-07 release, and the URL in the marker pack pins that dated filename rather than a "latest" pointer. If a future release changed the header format, an edit that did nothing would leave wrong taxonomy strings (leftover accession or [id: text read as a rank value, or a kingdom written as a domain) without any error. So each edit is checked: a header that does not start with >ACCESSION Root;, that still contains [id: afterwards, or whose first rank is then anything but k__Bacteria or k__Archaea (a kingdom such as k__Pseudomonadati, or a rank already written d__), stops the run and names the offending line. The log records how many headers were rewritten. The reference check then confirms that every lineage of the converted file starts with d__Bacteria.

Amplicon length: recovered by in-silico PCR

The rpoB reference is full-length, so reading its sequence lengths would return the length of the gene (about 4 kb), not of the amplicon. The marker pack therefore sets probe_mode: pcr, the same path used for 16S and 18S: two passes of cutadapt against the reference, using the supplied primers. For rpoB a third pass follows.

Pass	What it does
1	-g file:fwd with --discard-untrimmed — keep references carrying the forward primer at the 5′ end, and trim it off
2	-a file:rev_rc with --discard-untrimmed — of those, keep the ones carrying the reverse complement of the reverse primer at the 3′ end, and trim it off
3	the forward primer again, required at its full length, with --discard-trimmed — leave out of the statistics the bodies that still hold a second forward-primer site

The reverse primer is reverse-complemented first (with seqtk seq -r) so it can be searched for at the 3′ end. What survives both passes is the set of in-silico amplicon bodies; their lengths become the probe distribution. With the Univ_rpoB_deg primers the recovered bodies are around 387 bp — the amplicon minus the two primers, which cutadapt trims off in the two passes above.

Pass 3 is set in the rpoB marker pack (probe_exclude_second_site: true). In 17% of the FROGS genes that both primers amplify, the forward primer also matches, with as many mismatches as the probe allows, a stretch about 920 bp upstream of its real site, and cutadapt keeps that earlier match. Without pass 3 the probe gives a median of 389 bp but a p95 of 1,332 bp, and p95 is the statistic expected_length: auto uses; with it, q1 is 386 bp, the median 386 bp and p95 395 bp. The left-out bodies stay in the amplicons file. See A primer that matches twice.

This is the opposite arrangement to gyrB, whose reference ships already primer-trimmed and is therefore measured directly. The difference is a property of the two databases, not of the two genes.

If the probe finds nothing

When no reference sequence carries both primers, the probe stops rather than reporting an empty distribution:

[amplicon_probe] ERROR: no reference sequences had both primers. Check primer
sequences against the reference DB orientation and primer error tolerance.

That points at the primer FASTAs — the wrong orientation for this reference, or a mismatch tolerance too tight for the primers in use. The probe log is <out_dir>/logs/amplicon_probe.log.

The probe only runs when amplicon.expected_length: auto. It writes its statistics, and the recovered amplicons, to files whose names encode the marker, the reference tag (which ends in _excl-second-site for pass 3), and a 12-character hash of both primer FASTAs:

refdb/cache/probe_rpoB_rpob_refseq_cc_20240707_excl-second-site_<primer_hash>.json
refdb/cache/probe_rpoB_rpob_refseq_cc_20240707_excl-second-site_<primer_hash>.amplicons.fa.gz

Changing primers changes the hash, so the probe reruns; keeping the same primers across runs reuses the cached result. The JSON then feeds two downstream steps: pick_trunclen, which uses the resolved expected length to enforce the forward/reverse overlap constraint on truncLen, and the ASV length filter, which builds its auto window from the probe's q1 and p95 plus length_filter.window_margin. See Amplicon length and truncation for both.

probe_length_stat.rpoB must always be present

Which statistic of the probe distribution becomes expected_length is chosen by amplicon.probe_length_stat.rpoB (the shipped template sets p95). The key is checked at start-up whether or not expected_length is auto; with no such entry, the run stops with:

[MetaFlux] amplicon.probe_length_stat is missing the 'rpoB' key that marker rpoB
needs. Add it, e.g.
    probe_length_stat:
      rpoB: p95

A manual expected_length (a single integer, or a [min, max] pair) skips the probe itself, but the entry must still be there.

Taxonomy: RDP only

The FROGS package targets FROGS' own classifier; MetaFlux uses the converted FASTA with DADA2's assignTaxonomy instead. There is no SINTAX-formatted build of this reference, so the marker pack leaves taxonomy_sintax_db empty and MetaFlux refuses method: sintax at start-up rather than failing later inside the classifier.

taxonomy.min_boot sets the bootstrap confidence minimum and taxonomy.try_rc decides whether reverse-complemented sequences are also tried. There is no separate addSpecies step: unlike 16S, where species come from a second SILVA file, the converted rpoB trainset is already labelled to species, so the classifier reaches the species rank directly. amplicon.seed fixes the bootstrap RNG, so RDP results are reproducible run to run, apart from DADA2's random tie-breaking between equally good reference lineages, which can move a bootstrap value between identical runs (see Troubleshooting).

The classifier writes seven rank columns, taken from the marker pack. The converted reference already carries d__/p__/… prefixes, so the pack sets prefix_style: embedded and MetaFlux emits each value verbatim rather than prepending its own:

Slot	Rank column	Prefix in the taxonomy string
1	Domain	d__
2	Phylum	p__
3	Class	c__
4	Order	o__
5	Family	f__
6	Genus	g__
7	Species	s__

There is no kingdom slot, so a k__ filter token stops the run at start-up, and so does a domain other than d__Bacteria (the reference holds bacteria only).

Names follow NCBI's taxonomy as it was when FROGS built this release (July 2024): the phyla carry their current names, p__Bacillota and p__Pseudomonadota, and NCBI had not yet added the prokaryote kingdoms. The species slot is an underscore-joined binomial, s__Staphylococcus_simiae. That matters when writing filter tokens: they must match the reference's spelling, not the older Firmicutes/Proteobacteria names.

These are NCBI names, not GTDB ones. Unlike the GTDB-based gyrB reference, rpoB genus names carry no GTDB suffixes: it is g__Pseudomonas, never g__Pseudomonas_E. The two markers sit on different taxonomic backbones — gyrB on GTDB r226, rpoB on NCBI RefSeq — which circumscribe genera and name species differently, so labels are not interchangeable between them.

No paralog tag, and no default taxon filter

gyrB's reference carries a gene tag that separates true gyrB from co-amplified paralogs such as parE. rpoB has no close paralog of that kind, so the FROGS RefSeq set contains rpoB and nothing else, there is no equivalent tag to filter on, and MetaFlux recommends no default keep/discard list for this marker.

Because every reference sequence is rpoB, an off-target or paralogous amplicon cannot show up as a distinct label: it is assigned to whichever rpoB lineage it most resembles. gyrB's tag makes that visible; for rpoB the checks are the ASV length window, the bootstrap confidences, and the fraction of ASVs left unassigned.

The filter is still available, and the usual reason to reach for it on an rpoB run is dropping lab or kit contaminants that a negative control revealed:

amplicon:
  taxonomy:
    filter:
      enabled: true
      keep:    []
      discard: [g__Ralstonia, g__Bradyrhizobium]

Tokens are matched against whole segments of the taxonomy string, so g__Ralstonia hits a genus named exactly that and never a substring elsewhere in the lineage. Use the exact rank-prefixed name as it appears in the taxonomy column of asv_table.txt. See Keeping and discarding taxa.

An empty list is not the same as a disabled filter

An empty keep or discard list does nothing, which is the recommended starting point for rpoB. enabled: false switches the whole filter off. Either way every ASV survives — but if a filter ever removes all of them, assign_taxonomy stops with an error instead of writing a header-only table, since that is nearly always a marker/config mismatch.

No region extractor

MetaFlux has no rpoB equivalent of Metaxa2 (16S) or ITSx (ITS), so there is nothing to trim ASVs down to a target region. amplicon.extraction.enabled: true is therefore forced off for rpoB with a warning on stderr:

[MetaFlux] warning: marker rpoB has no target-region extractor; ignoring
amplicon.extraction.enabled: true (no extraction step will run)

Setting enabled: false explicitly keeps the run log clean. The ASV length filter still runs either way, so off-target amplicons of the wrong size are still removed.

What the output looks like

The final tables are the same four files as for every other marker, under 06.taxonomy/ in the output directory:

File	Rows	Notes
asv_table.txt	ASV IDs	per-sample counts plus one taxonomy column
asv_table_seqs.txt	ASV sequences	the same table keyed by sequence instead of ID
taxon_seq_table.txt	ASV IDs	one column per rank (Domain … Species) plus the sequence; the cells hold the bare names (Pseudomonadota, Xenorhabdus_nematophila), while the taxonomy strings keep the prefixes
asv_table_unfiltered.txt	ASV IDs, every ASV that reached the filter	the counts and taxonomy before the taxon filter, plus metaxa2_origin (NA for rpoB) and filter_status

An rpoB taxonomy string reads like a 16S one on SILVA 138.2 (a domain and no kingdom), with NCBI names:

d__Bacteria;p__Pseudomonadota;c__Gammaproteobacteria;o__Pseudomonadales;f__Pseudomonadaceae;g__Pseudomonas;s__Pseudomonas_fluorescens

Ranks the classifier could not assign above min_boot are simply absent from the string, so a shorter string means a less confident assignment rather than a missing rank.

Related pages

• gyrB — the other bacterial protein-coding marker, pre-trimmed reference and a paralog tag, but likewise RDP-only
• Amplicon length and truncation
• Keeping and discarding taxa
• Configuration

References

21. FROGS rpoB reference databank (2024). Bacterial rpoB genes from complete and chromosome NCBI RefSeq genomes, build 20240707. INRAE Toulouse. https://web-genobioinfo.toulouse.inrae.fr/frogs_databanks/assignation/rpoB/

32. Ogier, J.-C., et al. (2019). rpoB, a promising marker for analyzing the diversity of bacterial communities by amplicon sequencing. BMC Microbiology. (rpoB primers Univ_rpoB_deg)

The numbering follows the full list on Citation and references.