A fast laboratory arrives, and gradient length becomes a number
Kestrel Bioanalytical begins contributing on a 15-minute gradient. Quantifying its offset turns a methodological argument into a measurement.
Kestrel Bioanalytical began contributing in 2025-Q3. It is the fastest and cheapest laboratory in the index and it runs a 15-minute gradient, which is below the index minimum for the acylated incretins. Publishing its data required first measuring what the short gradient does.
The measurement
Across paired same-lot comparisons, Kestrel's purity results read a mean +0.24 pp above the reference across all compounds, and +0.51 pp on the incretin class specifically. The class-specific figure is the important one: it is where mass-identical diastereomers co-elute into the main peak on a gradient that short, and the offset is an integration artefact rather than a calibration one.
| Laboratory | Records | Share of quarter | Median purity |
|---|---|---|---|
| Janoshik | 409 | 49.9% | 98.82 |
| PeptideMeter | 132 | 16.1% | 98.75 |
| Medutest | 109 | 13.3% | 98.85 |
| VendorInvestigate | 61 | 7.4% | 98.19 |
| Nordanalyt | 52 | 6.3% | 98.85 |
| Kestrel | 56 | 6.8% | 99.07 |
What the index did instead of discarding the data
- Kestrel incretin results are excluded from class-level purity aggregates and shown separately on compound pages.
- Every Kestrel record carries a method-limitation note naming the specific impurities its gradient does not resolve for that compound.
- The verification status is set to METHOD LIMITED, which is a description rather than a judgement.
The alternative — refusing the submissions — would have removed a substantial part of the coverage for several thinly-tested suppliers. A known and quantified bias is more useful than a missing observation, and the index prefers to publish with the correction stated.
Gradient normalisation
The same arrival forced the index to formalise retention-time normalisation, because a peak at 9.9 min on a 15-minute gradient and one at 21.2 min on a 35-minute gradient are the same compound and the index needs to say so. The model is linear in gradient duration with a duration-independent term for dead volume, it is published in full, and it is an approximation that degrades at very early and very late elution.
The short bioregulator class produces most of the resulting false retention flags. That is a stated limitation of the model rather than a finding about the material.