Validation

Primer and band size validation

Every primer in brucetyper/assays.py was checked against the sequence printed in the publication that describes its assay, and every expected band size was checked against the product the primer pair actually gives in real Brucella genomes.

Method

87 RefSeq genomes were run through bruce_ladder.py at its defaults (--mismatches 2, --three-prime 3, --tolerance 5): up to 10 per species for B. melitensis, B. abortus, B. suis, B. canis, B. ovis, B. ceti, B. pinnipedialis, plus all available B. neotomae and B. microti, plus the reference and vaccine strains 16M, 2308, S19, RB51, 1330 and the B. suis biovar references. The full list is in validation-genomes.tsv, which also records the call each genome received from each of the four assays, so every call count on this page can be checked against it. The rough flag is not one of its columns, so the counts under The rough flag on the panel are the one exception.

"Observed" below is the product the pair gives in those genomes; n counts the genomes showing the main size. A locus only amplifies in the species it was designed to mark, so n is well below 87 for the discriminating loci.

Primer sequences

All 40 primer entries in assays.py, 33 distinct sequences, match the published sequences exactly. No substitutions, no typos. Entries outnumber sequences because several are reused: the universal IS711 primer serves five AMOS loci, the BMEI1426_BMEI1427 pair serves both Suis-ladder and Bruce-ladder, and primer 3 serves both loci of the RB51 junction assay. The three sequences added later for that assay were read from Fig. 1 of Vemulapalli et al. 1999 and are covered under RB51 junction PCR below. Two apparent discrepancies were checked and resolved in favour of assays.py:

Primer	Printed elsewhere as	Resolution
BMEI0536 R	CGCAGGGGAAAACAGCTATAA	The GGGG form has no exact match in any genome; the form in assays.py matches and gives the published 450 bp. The other spelling is a typo in that secondary table.
BMEI1688 F	TCAATCGCGTGAACAATGCT	That form has no exact match in 16M; the form in assays.py matches and gives the published 278 bp.

One assay-level difference is not a typo and is described under Bruce-ladder below.

Bruce-ladder

Published sizes from García-Yoldi et al. 2006 / López-Goñi et al. 2008.

Locus	Published	In assays.py	Observed in silico	n	Note
BMEI0998_BMEI0997	1682	1682	1682	74	2526 bp in RB51, where IS711 has inserted into wboA
BMEI0535_BMEI0536	450	450 / 451	450 (50), 451 (14)	50	449–452 across strains; 1049–1675 in ceti/pinnipedialis
BMEII0843_BMEII0844	1071	1071	1071	71	1077 and 1094 in two canis
BMEII0428_BMEII0428	587	590	590	85	Published size is 3 bp low; 590 is unanimous across all 85
BR0953	272	272	272	37
BMEI0752	218	—	no product	0	See below
BMEII0987	152	152	152	79	151 in two ceti
BMEI1436_BMEI1435	794	not used	794	—	See below
BMEI1426_BMEI1427	not a Bruce-ladder locus	774	774	76	See below

The fourth primer pair is the Suis-ladder locus, not the Bruce-ladder one

Published Bruce-ladder uses BMEI1436F/BMEI1435R (ACGCAGACGACCTTCGGTAT / TTTATCCATCGCCCTGTCAC) for a 794 bp band. assays.py instead carries BMEI1426/BMEI1427 (774 bp), which is the Suis-ladder locus, and its bruce profiles encode 774 to match.

Both pairs were tested across the panel. The published pair gives exactly 794 bp, and the pair in use gives exactly 774 bp; neither amplifies in B. canis. Because the profiles are written around 774 consistently, calls come out right — but this is a substituted locus, not an adjusted size, and the assay as implemented is not the published Bruce-ladder at this position.

BMEI0752 never amplifies

The BMEI0752 pair yielded no product in any of the 87 genomes. The forward primer has one exact match per genome; the reverse primer, GATGTGGTAACGCACACCAA, has none. The genomes carry GATGTGGTAACGCACACCCG at that site — the published primer differs in its last two bases, at the 3' end, which --three-prime 3 requires to match exactly. Relaxing to --three-prime 0 recovers a 219 bp product, against a published 218.

This is not a transcription error in assays.py: the published tables print GATGTGGTAACGCACACCAA, and the genome-matching spelling appears in no paper indexed by Europe PMC. The published primer itself mismatches Brucella at its 3' end. The consequence is that Bruce-ladder here runs on seven of its eight loci, which is consistent with 218 bp being absent from every profile.

Vaccine strains

RB51 and S19 are called from the two bands that separate them from abortus, and were added after this validation:

Genotype	Profile	Separated from abortus by	Confirmed in
RB51	2526, 774, 590, 450, 152	1682 shifts to 2526, IS711 inside wboA	GCF_011801185.1
S19	1682, 774, 450, 152	the 590 bp BMEII0428 band is absent	GCF_000018725.1, GCF_038420195.1

Adding them changed three calls in the 87 genome panel, all from N/A to the right vaccine strain, and no other call. Strain 2308 and eight of the nine abortus field strains still call abortus, the ninth being LBAB009, which calls N/A for the unrelated reason given under The seven N/A, so the vaccine profiles do not shadow it. Note that 2308 is reported as abortus by this assay, unlike AMOS, which groups it with RB51.

Concordance with the labels

Every call the panel produced, against the species each genome is labelled with:

Labelled	n	Called
abortus	13	abortus 9, S19 2, RB51 1, N/A 1
melitensis	11	melitensis 9, suis_bv2/suis_bv1_3_4 1, canis 1
suis	14	suis_bv2/suis_bv1_3_4 12, N/A 2
canis	10	canis 8, N/A 2
ovis	10	ovis 10
ceti	10	pinnipedialis_ceti 10
pinnipedialis	8	pinnipedialis_ceti 8
neotomae	6	neotomae 4, N/A 2
microti	5	suis_bv2/suis_bv1_3_4 5

So 78 of the 87 agree with the label, counting the three vaccine strains as abortus and the two merges the assay cannot resolve, microti with suis bv 2 and ceti with pinnipedialis, as agreement. Two contradict the label, and both are right to. Seven are N/A.

Two of the eleven melitensis are mislabelled submissions

melitensis__GTC-3P1517__GCF_055388155.1 calls suis_bv2/suis_bv1_3_4 and melitensis__GTC-3P1521__GCF_055389085.1 calls canis. Neither is an assay error. Both are in the independently reported set on the Large scale validation page, where MLST-9, the Hoover et al. 2026 real-time PCR and Mash distances place them exactly where Bruce-ladder does. The labels are wrong and the calls are right, so the panel's two disagreements are both submissions rather than assay failures.

The rough flag on the panel

The rough column of calls.tsv reads the 1682 bp wboA band alone. Across the 87 genomes it flags 13:

Flagged	Genomes	Why
B. ovis	10 of 10	Naturally rough, no wboA product at all
RB51	1	IS711 inside wboA, band shifted to 2526 bp
LBAB009	1	~15 kb deletion removing the locus, see below
neotomae NCTC10070	1	False positive: a 49 contig draft, 2.94 Mb, ~370 kb short of a full genome. The locus is not in the assembly rather than not in the strain

And the one it misses:

Not flagged	Genomes	Why it matters
B. canis	10 of 10	False negative: B. canis is naturally rough, but keeps an intact 1682 bp band. Its roughness does not come from losing wboA, so a band-level rule cannot see it

no therefore does not mean smooth. It means this one band is where it should be. Of the two naturally rough species, the flag catches B. ovis and misses B. canis.

The seven N/A

Genome	Level	Why it matches no profile
abortus LBAB009	Complete	~15 kb deletion removing the O-antigen locus, so no 1682 bp band. In detail below
canis 2010009751, SVA13	Complete	The BMEII0843_BMEII0844 products the locus table above records at 1077 and 1094 bp instead of 1071. Both sit outside the 5 bp window, and the canis profile needs that band
neotomae NCTC10070	Contig	49 contig draft, 2.94 Mb, ~370 kb short of a full genome. The locus is not in the assembly
neotomae 5K33	Scaffold	Not traced
suis ZW043, ZW046	Complete	Not traced. Suis-ladder calls both bv01

Five of the seven are complete genomes, so fragmentation is not the general explanation, and three are not traced to a cause. Of the four that are, only LBAB009 is a property of the strain. The two canis are a --tolerance artefact — a real product 6 and 23 bp off a canonical band — and NCTC10070 is an assembly gap. The same tolerance trade-off is measured over 2051 genomes on the Large scale validation page, where widening the window is shown to cost more than it recovers.

LBAB009 in detail

abortus__LBAB009__GCF_030761395.1 gives 774, 590, 450 and 152 but no 1682 bp wboA band.

It is not an assembly artefact: the genome is complete, two chromosomes, 3.29 Mb. Neither BMEI0998_BMEI0997 primer has any match in it, because roughly 15 kb of chromosome 1 is missing, corresponding to 2308 NC_007618.1 953,840..968,920. That block carries BAB_RS20745 and BAB_RS20750, the two glycosyltransferases of the O-antigen locus (wboA, wboB), so the strain is most likely a naturally rough isolate. Its sibling isolates LBAB008 and LBAB016 both carry the band, and AMOS calls LBAB009 abortus_bv1_2_4 normally.

N/A is the right answer here. The strain is B. abortus carrying a real deletion, and no profile was added for it: calling it abortus would hide the finding, and a profile fitted to one isolate would not generalise. It differs from RB51, where IS711 inserts into wboA and shifts the band to 2526 bp rather than removing it.

Suis-ladder

Published sizes from López-Goñi et al. 2011.

Locus	Published	In assays.py	Observed in silico	n	Note
BMEI1426_BMEI1427	774	774	774	76
BR1080	197	197	197	43
BMEI1688_MEI1687	278	278	278	57
BMEI0205	299 / 425 / 551 / 614	299 / 425 / 551 / 614	299 (19), 614 (13), 551 (12), 425 (11)	55	VNTR; also 173 (11, ovis), 236 (9, abortus/melitensis), 803 (5, microti), 549 (3)

All five biovar reference strains were called correctly: bv01 through bv05.

AMOS

Published sizes from Bricker and Halling 1994 and 1995.

Locus	Published	In assays.py	Observed in silico	n	Note
abortus_IS711	498	494	494	10	Published size is 4 bp high; 494 in all 10 abortus, S19 and RB51 included
melitensis_IS711	731	733	732 (8), 733 (2)	10	Published size is 1–2 bp low; 733 is the 16M reference value, 732 the field-strain value. Both inside --tolerance
ovis_IS711	976	976	976	9
suis_IS711	285	285	285 (2), 276 (1)	3	285 confirmed in 1330; amplifies in very few genomes
RB51_2308_IS711	364	387	387	10	Published size is 23 bp low. 387 in 2308 and RB51, and in 7 ovis and 1 pinnipedialis, which carry their own tandem IS711 pair. 2112 bp in three ceti
eri	178	178	178	85	Present in every genome but S19

RB51 junction PCR

Primers from Vemulapalli et al. 1999, Fig. 1. Validated over the same 87 genomes with --rb51.

The three primers

	Sequence	Role
Primer 3	GCCAACCAACCCAAATGCTCACAA	Forward, in wboA upstream of the insertion point. Shared by both pairs
Primer 1	TTAAGCGCTGATGCCATTTCCTTCAC	Reverse, in wboA downstream of the insertion point
Primer 2	TTTAGTTTGCCGTAATATAGGT + CTAGAACCTGTC	Reverse, chimeric: 22 bases of wboA fused to 12 bases of IS711

Primer 2 is what makes the assay specific. Its wboA half has an exact site in each of the 75 genomes that carry the locus and in none of the other 12. Its IS711 half is the terminal 12 bp of the element's right end, not an internal stretch, so its site count does not track IS711 copy number: one to three exact sites per genome, 14 genomes having only one, and in 2308 only one of its two sites is a real IS711 end, the other a coincidental 12-mer. The fused 34-mer has an exact site in RB51 and in no other genome. It cannot anneal unless IS711 is inside wboA.

The whole margin lives in primer 2's 3' terminal 8 bases. In every wboA bearing non-RB51 genome the target reads TTCAAGCG|CTAG|ACCTATATTACGGCAAACTAAA against the primer's GACAGGTT|CTAG|ACCTATATTACGGCAAACTAAA: the 22 base wboA half and the first 4 bases of the IS711 half already match, and all five mismatches fall in the last 8 bases, the 3' terminal base included. The mismatch budget and the 3' anchor therefore block it independently.

The insertion

Aligning the RB51 and 2308 wboA products places the insertion 342 bp into the Bruce-ladder amplicon, at RB51 NZ_CP046720.1:964,987, spanning 964,987–965,830. The insertion adds 844 bp, which is the whole difference between the two sizes of the wboA locus below. IS711 is usually quoted as 842 bp in the literature; 844 bp is what these assemblies carry, and a second full length copy in 2308 (NC_007618.1:541,972) measures 844 bp as well.

The breakpoint is ambiguous over a 4 bp window from sequence alone, because the wboA target and the element both end in CTAG: a naive common prefix and suffix walk returns 964,983 and an 840 bp insert. 964,987 is the placement at which the insert begins with the conserved IS711 left end GGCGTGTCTGCA, which is what fixes it.

Band sizes

Locus	Primers	In assays.py	Observed in silico	n	Note
wboA	3 + 1	456	456 (74), 1300 (1)	74	1300 bp is RB51, 456 + the 844 bp insertion. It is deliberately not a profile size: it is tolerated as an extra band so that an RB51 draft missing it still calls RB51. No product in the 12 genomes without the locus
wboA_IS711	3 + 2	919	919	1	RB51 only, 0 mismatches on both primers. No product in the other 86 genomes, 2308 and S19 included

The sizes Vemulapalli et al. print for these products are not recorded here, unlike every other assay in this file: the figure gives the primer sequences but no sizes, and the sizes quoted for this assay in secondary sources could not be reconciled with any genome, so only the observed values are stated. See Two sets of primers that are not this assay.

Calls over the 87 genomes

Genomes	Call	Why
RB51	RB51	919 bp junction band, plus 1300 bp in extra_bands
74 others	not_RB51	wboA intact at 456 bp, no junction band
10 B. ovis	N/A	Naturally no wboA product at all
abortus LBAB009	N/A	O-antigen locus deleted, see above
neotomae NCTC10070	N/A	49 contig draft that does not contain the locus

No false positive, and every call the assay declined to make was a genome with no wboA locus to read rather than a silent failure.

What the mismatch budget is carrying

The N/A set is the set of genomes in which primer 3 has no site within the tolerated mismatch budget, which is not the same as having no exact site. Primer 3 has no exact site in 19 of the 87 genomes: the 12 above, plus seven B. suis biovar 2 genomes that carry GCCAACCACCCCAAATGCTCACAA, an A→C transversion at primer position 9. Primer 1 matches those seven exactly, so at the default --mismatches 2 they give a clean 456 bp product and call not_RB51; at --mismatches 0 they fall to N/A and the panel splits 19 / 67 / 1 instead of 12 / 74 / 1. The SNP sits far from the 3' end, so the --three-prime anchor does not block it.

In the other direction the assay has room to spare. The next best placement of primer 2 anywhere in the panel, on either strand and with no 3' anchoring, is 5 mismatches, so the calls are unchanged from --mismatches 2 through --mismatches 6. That 5 mismatch site is real and sits 75 bp from the primer 3 site in all 74 wboA bearing non-RB51 genomes — it is the intended target minus the element. Forced through with --mismatches 8 --three-prime 0, it yields a 75 bp product in all 74, and the calls still come out 1 / 74 / 12, because 75 bp is nowhere near the 919 bp band.

Two sets of primers that are not this assay

Both were checked against the genomes and rejected, which is why the sequences above were taken from the figure rather than from the surrounding text:

Sequences	Verdict
TTATCGCAATCCATCGCGAT, AGTGCGACGGAGAGTCCCAA, quoted for this assay by secondary sources	Not Brucella. The best placement of either anywhere in RB51 or 2308 is 4 mismatches, and 6 or more in the O-antigen region itself
GGATGTCGACCAGCCCTCCACATCAATAGC, TTGCGGATCCTTTACTCGTCCGTCTCTTAC, from the paper's methods section	Genuine but not diagnostic. They match exactly once their engineered SalI and BamHI tails are set aside, but they are the cloning primers used to characterise the insertion: they flank the whole gene and give 2114 bp in 2308 against 2958 bp in RB51, both strains amplifying

Summary of differences

Sizes in assays.py that deliberately depart from the published figure, each confirmed by the genomes:

Assay	Locus	Published	Used	Difference
amos	RB51_2308_IS711	364	387	+23
amos	abortus_IS711	498	494	−4
bruce	BMEII0428_BMEII0428	587	590	+3
amos	melitensis_IS711	731	733	+2

Open points, neither of which changes a call today:

• BMEI1426_BMEI1427 (774 bp) stands where published Bruce-ladder uses BMEI1436_BMEI1435 (794 bp).
• BMEI0752 cannot amplify at the default 3' anchor, so the published 218 bp band is unreachable.