NextGenAMR — a product by Vanguard Biotech Systems
Evidence and validation
This page documents the scope, methodology, status and publicly approved results of the NextGenAMR evaluation programme. It publishes what is approved — and clearly states what is not yet available.
Performance results for this public version have not yet been reported.
NextGenAMR is currently undergoing controlled evaluation and internal, non-clinical validation. Validation results have not been published and must not be inferred.
Genomic evidence and the reference standard
NextGenAMR organises resistance-related genomic evidence within its published scope. Evaluation compares predefined product outputs against a reference standard specified by a protocol. Detecting a marker does not by itself equal a clinical category, and the absence of a marker does not guarantee susceptibility.
For each antibiotic in scope, the model output is an estimate of the calibrated probability of resistance — never a categorical susceptible/intermediate/resistant (S/I/R) result.
NextGenAMR does not replace phenotypic antimicrobial susceptibility testing (AST). Genomic outputs must be interpreted together with AST, microbiological context and applicable procedures.
What is evaluated
- Operational species
- Escherichia coli
- Evaluated output
- Calibrated probability of resistance
- Scope version
- 2026.07.0 · 2026-07-15
Antibiotic panel (9)
- Amoxicillin–clavulanic acid
- Ampicillin
- Ciprofloxacin
- Ceftazidime
- Gentamicin
- Meropenem
- Nitrofurantoin
- Piperacillin–tazobactam
- Trimethoprim–sulfamethoxazole
Methodology, work in progress & results
A validation protocol has been drafted: it defines the intended reference-standard comparison, the endpoints, and the handling of discordances, abstentions and failures. Its scientific parameters and its publication are not yet approved.
Fixing and freezing the interpretive standard, assembling an independent validation cohort, and obtaining scientific and publication approval. No cohort has been assembled and no results have been computed for publication.
Not yet publicly reported.
Public protocol
- Intended use evaluated
- Whether NextGenAMR's per-antibiotic outputs, within its published Escherichia coli scope, agree with a reference standard under a defined protocol — for professional review, not autonomous clinical use.
- Design
- A retrospective comparison of predefined product outputs against a reference standard on an independent set of bacterial isolates, reported per organism and antibiotic.
- Cohort type
- An independent set of Escherichia coli isolates with paired reference-standard results, independent from model training and calibration.
- Input
- Whole-genome sequencing (paired-end FASTQ) of cultured isolates within scope.
- Preconditions
- Isolates and reference-standard results that meet the protocol's quality and independence criteria.
- Reference standard
- Phenotypic antimicrobial susceptibility testing under a recognised interpretive standard. The exact standard and version will be fixed and frozen in the approved protocol.
- Inclusion criteria
- Isolates within the published scope with a valid, independent reference-standard result. A minimum number of cases per organism and antibiotic will be required by the approved protocol.
- Exclusion criteria
- Cases failing quality control, with species-confirmation mismatch, or with any overlap with model training or calibration data.
- Discordance handling
- Discordances are analysed, not hidden; intermediate reference results are not collapsed into susceptible or resistant.
- Abstention handling
- Abstentions are reported separately and are not scored as correct or incorrect.
- Failure handling
- Technical run failures are reported as such and separated from scientific outcomes.
- Analysis rules
- Results are reported per organism and antibiotic. A global figure never replaces per-antibiotic reporting.
Intended endpoints
- · Categorical agreement (once an approved output↔category rule exists)
- · Major error rate
- · Very major error rate
- · Abstention rate (reported, not scored)
Protocol limitations
- · The protocol is a draft; its scientific parameters and publication are not yet approved.
- · The product emits calibrated probabilities, not S/I/R categories; categorical agreement requires an approved output↔category correspondence that does not yet exist.
Metric definitions
Prepared definitions, grouped by family. They are pending scientific review; publishing a metric definition is not the same as reporting a result.
Scientific metrics
- Categorical agreementDefinition pending scientific review
- Within evaluable cases and under a defined protocol, the proportion of cases whose evaluated output category matches the reference-standard category. It is not a synonym for “accuracy”, and it requires an approved rule mapping the product's calibrated probability to a category — which does not yet exist.
- Major errorDefinition pending scientific review
- A resistant result emitted by the system when the reference standard classifies the case as susceptible, over the approved denominator. The exact formula, handling of intermediate categories and denominator come from the adopted protocol and standard.
- Very major errorDefinition pending scientific review
- A susceptible result emitted by the system when the reference standard classifies the case as resistant, over the approved denominator. The exact formula, handling of intermediate categories and denominator come from the adopted protocol and standard.
Operational metrics
- Abstention rateDefinition pending scientific review
- The frequency with which the system abstains from emitting the evaluated output due to foreseen conditions — insufficient evidence, inadequate quality, or output out of scope. Abstention is not an error, a success, or a technical failure.
- Failed-run rateDefinition pending scientific review
- The proportion of initiated runs that do not produce the expected technical output because of a failure of the process, infrastructure, a dependency or the input, according to the protocol.
- QC exclusion rateDefinition pending scientific review
- The proportion of inputs excluded by quality control before evaluation, according to the protocol.
System metrics
- Bioinformatic runtimeDefinition pending scientific review
- Bioinformatic time measured between defined technical points. It is not clinical turnaround, time-to-treatment, total laboratory time, or an operational saving. Any future figure must state the execution environment, compute profile, input characteristics, number of runs and statistical distribution.
Results
No clinical performance metrics (sensitivity, specificity, categorical agreement, VME/ME, accuracy) have been published. Performance evaluation is in progress.
Not yet publicly reported
No performance results are publicly reported for this version. When results are approved, they will appear here disaggregated by organism and antibiotic, each with numerator and denominator, cohort, reference standard, product and panel version, cut-off date, measurement conditions and an interval — never as a single global figure.
Scope & evidence by organism
Escherichia coliOperationalNo published results
- Public status
- Operational scope, internal non-clinical validation. No performance results reported.
- Intended use
- Organise resistance-related genomic evidence for the published antibiotic panel and produce calibrated, reviewable outputs for professional interpretation.
- Out of scope
- · Autonomous diagnosis or treatment selection.
- · A replacement for phenotypic AST.
- · Use on species or antibiotics outside the published scope.
- Inputs
- Whole-genome sequencing (paired-end FASTQ) of a cultured Escherichia coli isolate.
- Preconditions
- A single cultured isolate meeting quality thresholds; not a primary sample and not metagenomics.
- Outputs
- · A calibrated probability of resistance per antibiotic (never S/I/R).
- · Structured JSON and a PDF report for professional review.
- · Run provenance (tool, database and model versions, hashes).
- Components (high level)
- A six-model calibrated ensemble per antibiotic over genomic evidence. Internal design (model families, features, thresholds, calibration) is not published.
- Quality & abstention
- Read and genome quality gates apply; the system abstains on species mismatch, low confidence, insufficient evidence or out-of-scope input.
- Known limits
- · No clinical performance results are published for this version.
- · Generalisation beyond the evaluated conditions is not established.
- · Outputs depend on input quality and require professional review.
- Published results
- Not yet publicly reported
- Version / review
- v1.0.0 · 2026-07-18
Demonstrations and their limits
The demonstration shown on the product pages is generated from an actual system run on a real Escherichia coli sample. It illustrates the product's outputs; it is not, by itself, a validation result, a performance metric or a clinical report.
Demonstration generated from real bacterial sample data during internal testing with the current version of NextGenAMR. The content illustrates product outputs and does not by itself constitute a clinical validation result or treatment recommendation.
How it works →Open limits
- No clinical performance results are publicly reported for this version.
- The published scope is a single species (Escherichia coli) with a closed antibiotic panel; other species are in internal development and are not publishable.
- The product emits calibrated probabilities, not S/I/R categories, and does not replace phenotypic AST.
- Results depend on the reference standard and interpretive breakpoints ultimately adopted, which can change over time.
- The system may abstain, and outputs outside the published scope are not produced; all outputs require professional review.
Open scientific questions
- Which interpretive standard (and version) will be fixed and frozen for the validation protocol.
- The composition, size and source of the independent validation cohort.
- Whether an approved rule will map the product's calibrated probability to a category for categorical-agreement analysis.
- Publication authorization for any future result.
Versioning & history
Content version v1.0.0 · Last updated 2026-07-18
- 2026-07-18v1.0.0First public evidence & validation page. Programme status: validation in preparation. No performance results reported.
Discuss a validation collaboration
If your institution runs phenotypic AST and would consider a scientific validation collaboration, we would be glad to discuss it. The contact form does not accept samples, sequences, genomic files, sample identifiers or patient data.