PeptideMeter (revision 19)

PeptideMeter is a public database that collects submitted peptide test results and aggregates them into tallies by supplier and by compound, with individual entries traceable to the report they came from. It is one of the four testing and review services referred to throughout this wiki.^[1]

A database of this kind produces a different kind of object from an analytical laboratory. Janoshik Analytical and Medutest produce determinations on samples; PeptideMeter produces statistics over submissions. Treating an aggregate figure as though it were a measurement, or a measurement as though it were representative, are the two errors this distinction is meant to prevent.^[1]

Because the unit of record is a submitted report rather than a purchase or a lot, every tally derived from it is subject to self-selection at two stages: who commissions testing, and who submits the result.

An aggregate over submissions is a statistic about the submissions, not about the supplier's output. If ten reports on a supplier's material show a mean purity of 98.6%, that is a fact about ten samples that somebody chose to test and chose to submit.^[1]

The spread matters more than the mean. A supplier with ten submissions clustered within half a percentage point is characterised differently from one with ten submissions spanning five points at the same mean, and a bare mean conceals the difference. Where a distribution is available it is more informative than any single number derived from it, since the attributes a peptide specification covers are themselves multiple.^[2]

Sample size is the other axis. A tally of three submissions and a tally of ninety are not comparable, and ranking suppliers by mean purity without regard to submission count produces orderings driven mostly by noise. This is a general property of small-sample rankings rather than a criticism of any particular tally.^[3]

Two filters operate before a result reaches a database of this kind.

1. Who tests. Purchasers who commission independent analysis differ systematically from those who do not — they are more experienced, buy more, and often buy from suppliers they already trust.
2. Who submits. A favourable result is more likely to be submitted by a satisfied purchaser than an unfavourable one is by a dissatisfied one, though the reverse also occurs when a purchaser is aggrieved.

The net direction of the bias is not determinable from the data itself, which is why a tally should be read as a lower bound on variability rather than as an estimate of central tendency.^[1]

The same caution applies to this wiki's own community tally, which is labelled self-reported throughout and is not presented as a survey.

Twenty organisations have articles on this wiki. Their articles report reception tallies drawn from submissions to this project, not from PeptideMeter or any other external database, and the two are not expected to agree because the populations differ.^[1]

Cross-citing figures between communities is a recurring source of confusion: a figure computed over one set of self-selected submissions is not comparable with a figure computed over a different set, even when both are described as "purity" and both are expressed as percentages — the underlying chromatographic determinations need not have been made on comparable systems.^[4]

• Third-party testing
• Janoshik Analytical
• Medutest
• VendorInvestigate
• Purity claim inflation