R1Standard curve fit
Weighted four-parameter logistic fit with back-calculated recovery per calibrator. R² is reported as a response-space diagnostic and is never the acceptance criterion.
Gate · back-calculated recovery within tolerance
Research Use Only · Analytical quality control
Know whether the run was valid — and be able to show how you know.
Four routines evaluate calibration, replicate precision, amplification quality and batch drift. Each records the acceptance parameters it was given, the result it produced, and who ran it. Nothing is assumed on your behalf.
R1 · recorded run · 9-point calibrator series
The same curve, judged two ways.
Accepted
R² 1.000000
0.0% deviation
Every calibrator back-calculates to its nominal concentration.
Rejected
R² 0.999953
11.1% deviation
Worst calibrator recovery deviates 11.1% (limit 5.0%).
An R² of 0.999953 clears any conventional 0.99 gate. The same curve back-calculates its top calibrator well outside tolerance. Recovery is the acceptance gate; R² is a diagnostic.
The build reports its own state
Measured 2026-10-06
527Tests passing0 failing.
97.0%Mutation score97 of 100 injected faults caught by a test.
96.54%Line coverageOf 3,121 statements. Reported, not gated.
4 / 4Quality gatesCoverage, module floor, mutation, failing tests.
All four quality gates pass. Mutation score is the gate; coverage is reported, not gated. Coverage measures execution, not assertion, and gating on it rewards tests that assert nothing. Reproduce every figure above with
python measure_quality.py --mutation.Four routines
Each one refuses to run until it has been told what acceptable means.
R2Replicate precision
Per-marker mean, standard deviation and %CV, with the Grubbs outlier policy applied exactly as configured. Every verdict names the marker and the reason.
Gate · %CV within the configured limit
R3Amplification QC
Ct calling, amplification efficiency, no-template-control evaluation and ΔΔCt normalisation. Refuses to run until the assay-specific protocol parameters are supplied.
Gate · protocol parameters present, NTC policy met
R4Batch drift
Westgard multirules against the laboratory’s own established control limits, with zone concordance and completeness. The batch being judged never defines the limits used to judge it.
Gate · no rule violation against established limits
What it will not do
A result the software cannot stand behind is withheld, not estimated.
Most of the engineering is in the refusals. They are listed here because a laboratory evaluating this needs the boundary before the capability.
No scientific defaults. R3 and R4 stop and name the missing parameter rather than assuming a reasonable one.
Censored values carrying < or > are excluded and counted, never coerced to the number beside the symbol.
Blank cells are excluded and counted separately. An empty cell is not a zero.
Columns that fall or rise systematically are flagged as a possible dilution series rather than analysed as replicates.
Absent evidence is reported as absent. A missing measurement never reads as a passing one.
Evaluation pathway
From a technical review to a controlled pilot, in four steps.
01
Technical reviewInspect the API, the evidence record and the refusal behaviour.02
Instrument mappingA de-identified native export from your own instrument. No PHI.03
Protocol lockYour assay-specific R3 and R4 acceptance parameters, recorded.04
Pilot environmentSeparated data, backups and a tested restore before any live run.Evaluation copies run against de-identified native exports. No patient data, and no specimen, is required at any point in this pathway.
reforrest@omnicareoasis.help