Skip to content
Methods

Reference checks

Reference checks: Drybench Protein

This page lists the reference checks published for Drybench Protein. Each set gives the methods it covers, what the expected values were computed with, the number of cases, the tolerance the app must meet, the date, and the file holding the inputs and expected outputs, so you can open a case and compare it with what the app gives.

Every case was chosen and recorded by the Drybench team; none has been independently reviewed. A reference check tests the calculation, not your data, your experiment or whether a model suits it. Methods without a set are listed on the methods page with their level.

Methods and levelsDocumentation

Published sets

Published verification sets

MethodsOracleFixturesToleranceDateFile
protein.residue-mapBiopython47exact partners, states, null masks, refusals and allocation bounds2026-10-23residue-map-v2.jsonRecorded for version 0.5.1; the test suite checks the current engine against it.
protein.residue-mapPython standard library8exact partners, states, null masks and doubled scores, counts as decimal strings; above 20,000 optima only the counting reference and Biopython's optimal score apply; the 47 v2 cases reproduced2026-10-24residue-map-v3.jsonRecorded for version 0.8.0; the test suite checks the current engine against it.
protein.residue-map, protein.source-confidenceBiopython102exact parser, bands, holes, counts, targets and binding; means within 1e-122026-10-23confidence-v1.jsonRecorded for version 0.6.0; the test suite checks the current engine against it.
protein.structure-evidence-capturePython standard library13exact: seven byte identities, three PNG cases, three scene identities2026-10-18evidence-v2.jsonRecorded for version 0.3.0; the test suite checks the current engine against it.
protein.structure-evidence-capturePython standard library1exact sequence comparison; renderer application is a separate check2026-10-18evidence-v3.jsonRecorded for version 0.3.0; the test suite checks the current engine against it.
protein-average-mass, extinction-280-paceBiopython6designed constructs: ε280 and the aliphatic index exact, MW within 1 Da of Biopython, assembled sequences, the C-terminal refusal code and the added initiator Met exact2026-10-20construct-cropper-v2.jsonRecorded for version 0.4.0; the test suite checks the current engine against it.
protein-average-mass, extinction-280-pace, isoelectric-point-bjellqvistBiopython10pI within 0.01 of the reference, which clears Biopython's terminal-aspartate and -glutamate override and bisects pH 0 to 14 in place of its 4.05 to 12, with no case returning a bound; the two untagged convention cases also within 0.01 of ExPASy ProtParam's printed value for the same sequence, recorded 5 October 2026, so they carry a second oracle; ε280 and the aliphatic index exact; MW within 1 Da; assembled sequences exact2026-11-07construct-cropper-v3.jsonRecorded for version 0.11.0; the test suite checks the current engine against it.

Oracles and build-time tools

Software the app's numbers were checked against, or its data built with. None of it ships with the app or is a dependency of it.

  • Python standard library 3.9.6 (PSF-2.0; Oracle; 2026-10-18). Retained-image byte, PNG, scene and selection references
  • Python standard library 3.13.17 (PSF-2.0; Oracle; 2026-10-24). Exact-integer optimal-path counting, the residue map's second, non-enumerating reference (with Biopython 1.88 for BLOSUM62 and its optimal score); never shipped
  • Biopython 1.88 (Biopython License Agreement (BSD-3-Clause file by file); Oracle; 2026-10-23). ProtParam construct properties; all-co-optimal protein correspondence and exact confidence projection; never shipped