FILE
DL–2026–07
METHOD
vtMG · preprint
CLASS.
TECHNICAL · PUBLIC
K FACTOR
prior + posterior
PAIRS
152 with a time axis
CFUeq/g
the unit the standard is written in, and the unit your culture archive is written in
8
regulatory indicator groups, each with its factor by matrix and treatment
126+
food taxa with ribosomal copy number curated from rrnDB and NCBI
152
metagenomics-and-culture pairs with a real time axis, from 16 to 117 days after manufacture
+
+
+
+
What makes it unique
It is the only system that carries a metagenomic read all the way to the reportable unit with its own wet-lab method behind it: differential pre-lysis and DNase digestion before sequencing, so that free DNA and DNA from membrane-compromised cells do not enter the count. What comes out is a culture equivalent, not a proportion dressed up as a figure.
And the conversion factor is no invented constant: it is an empirical prior with its mechanism and its reference, which the engine fuses with real metagenomics-and-culture pairs by Bayesian shrinkage. With no data of your own the prior rules; as pairs arrive, the prior washes itself out. The prior is independent of the validation set — a written anti-circularity rule — and the metrics stay out of sample.
01
The state of the art
A percentage cannot be reported. And it does not compare with your archive.
Sequencing delivers proportions; regulation demands counts per gram, and twenty years of laboratory archive are in plate units. Without conversion, the new datum speaks neither to the standard nor to the past.
There is also a problem selective media cannot solve: they count presumptive groups that overlap. Enterobacteriaceae contains coliforms, which contain E. coli — the same isolate is counted three times. An enterococcus falls at once into lactic acid bacteria and into its own count. Resolving to species undoes that double counting.
02
The wet-lab method
A number without a protocol is not interpretable.
The literature says spores are under-detected. The live data said the opposite. There was no contradiction: there were two extraction protocols, and which one was in play had not been declared.
Standard extraction
The spore lyses poorly and its DNA is lost
Under bead beating the spore resists, its material never reaches the sequencer, and the molecular route sees less than there is. That is what the classical literature describes, and it is true under that protocol.
Negative bias · under-detection
vtMG · our own method
Differential pre-lysis and DNase digestion
Free DNA and DNA from membrane-compromised cells are digested before sequencing: what remains is signal from viable cells. The spore, impermeable to DNase, survives the filter and is read in excess — and that is declared.
Positive bias · spore over-detection
And the bias is not a constant: it is a trajectory
At early time points, what the molecular route sees and culture does not are spores and cells sublethally injured by the treatment. As the spore germinates and the bacterium recovers, culture picks it up and the two routes converge. All four indicators decay monotonically: none of them grows. A growing bias would contradict the mechanism, which is why the validator rejects it.
The time axis is real and already in the database: days between manufacture and assay. There is no need to invent it.
FIG. 01
Molecular excess in log₁₀ · real series T1→T3
Indicator
T1
T2
T3
Sulphite-reducers
+5,88
+2,94
+1,81
Enterobacteriaceae
+4,60
+2,63
+1,57
Aerobes
+2,07
−0,47
−1,20
Lactic acid bacteria
+2,43
−0,89
−1,55
MIGAS BACALAO MAP series. Positive = the molecular route reads high.
The honest limit of the fit
The trajectory is clean within a series and dilutes when studies are pooled: over 152 pairs, the decay rate is confounded with matrix and treatment — R² 0.33 for aerobes, 0.00 for Enterobacteriaceae. That is why mechanism and sign enter as curated while the numbers travel as pending, with their n and their R² in plain sight.
Prudence is applied before concluding: pairs where both methods sit at their floor, or where culture only provides a bound, are discarded, and at least three quantifiable pairs and one log of deviation are required. The measured effect was going from six or seven flagged indicators per analysis to one or two; with a single pair per indicator, none is issued.
03
How it works
Three steps, and a factor that learns.
From percentage to reportable unit, with the critical distinction between amplicon and shotgun: the first corrects copy number only; the second, copy number by genome size.
01
Copy number correction
Abundance is normalised by the ribosomal gene copies each genome carries, against a curated base of more than one hundred and twenty-six food taxa. It applies to amplicon only: in shotgun the correction is copy number by genome size, and confusing the two falsifies the result.
02
Anchoring to DNA mass
The corrected abundance is anchored to total DNA mass measured by fluorimetry, which yields cells per gram. That is the step that turns a proportion into a quantity.
03
Calibration by matrix and treatment
The factor is applied by indicator group, food matrix and treatment — high pressure, thermal, raw — sample by sample, not per analysis. The result is reportable CFUeq/g.
04
The factor updates with your data
The posterior is the strength-weighted mean of the prior and the real pairs: with no data of your own the prior rules, and as metagenomics-and-culture pairs accumulate the prior dilutes. The prior is never deactivated, so that both can fuse.
05
Comparison and validation
Against paired culture, Lin's concordance, Passing-Bablok, Bland-Altman and temporal drift are computed. A value below the detection limit is replaced by the limit itself — the EFSA and WHO prudence convention — not by half of it, which is an analytical-chemistry criterion, nor by exclusion, which biases downwards.
The eight indicator groups
Aerobes
Enterobacteriaceae
Coliforms
E. coli
Sulphite-reducers
Yeasts and moulds
Lactic acid bacteria
Listeria
04
The contract
What goes in, what comes out, what it chains to.
In
Abundances from OmniOta or your own table, total DNA mass by fluorimetry, and matrix and treatment declared sample by sample. The culture archive is imported from a spreadsheet, with below-limit values detected automatically.
Out
Per sample and indicator group: CFUeq/g and its logarithm, the factor applied, contributing taxa with their breakdown, compliance against the criterion, the paired culture reference, the discrepancy level and a confidence score.
Chains to
It receives the profiles from OmniOta and feeds FoodSpan, which needs the quantified time series to fit spoilage kinetics and compute shelf life.
05
Technical sheet
Name
Deep Estimation of Live Organisms by Read Equivalence to Agar Numbers.
Input
Amplicon or shotgun abundances from OmniOta or your own table, DNA mass, matrix and treatment per sample, and optionally the culture archive imported from a spreadsheet.
Output
CFUeq/g and log₁₀ per sample and indicator group, with the factor applied, contributing taxa, compliance against the criterion, classified discrepancy and a confidence score.
Conversion factor
A curated prior by group × matrix × treatment with its value, its interquartile range, the strength behind it, the biological mechanism and the reference; fused with real pairs by Bayesian shrinkage.
Censored data
A "<10" is replaced by 10. Not by half the limit — the analytical-chemistry convention — nor by exclusion, which biases the whole set downwards.
Anti-circularity
The prior is independent of the validation set and is not recalibrated with it. Metrics stay out of sample, and this is written down as a project rule.
Limit · unit
CFUeq/g is a calibrated culture equivalent, not a plate count: it includes the viable but non-culturable, which is exactly what the plate cannot see. For formal regulatory validation it is complemented with culture to ISO standard.
Limit · spores
Under the viability method the spore is impermeable to DNase and its DNA survives the filter: the report gives the sign and the magnitude of the excess instead of a clean, false number.
Limit · curation
Only mechanism and sign enter as curated. The trajectory numbers travel as pending with their n and their R², awaiting more series with culture in order to stratify by matrix.
Our own wet-lab method, in preprint. Copy-number base curated from rrnDB and NCBI. EFSA/WHO censored-data convention. Regulation (EC) 2073/2005 and ICMSF criteria for the compliance check. Statistical validation: Lin's concordance, Passing-Bablok, Bland-Altman and temporal drift detection.
The unit matters. Without it, the datum never enters the dossier.
BIOTECNO.org · Vitoria-Gasteiz
EU + Codex