Data

The data behind the atlas, as versioned files you can download and reuse. Each file is generated from the same sources the site reads when the site is built, so the two cannot disagree. Each one states its version, date and licence, and each has a JSON Schema beside it. This is release v4.6, last updated November 4, 2026. For research use only; scores and recommendations reflect the authors' assessment of published literature.

Files

Assessed platforms

Specifications, strengths and limitations for the six assessed platforms.

platforms.json Schema

Platform catalogue

The 46-platform catalogue, with evaluation level and integration readiness.

all_platforms.json Schema

Scores

The ten scored dimensions, each platform's scores, and which platforms are flagged lower confidence.

scoring.json Schema

Evidence base

Studies with their key findings as claim objects: text, platforms, topics, figures and source location.

evidence.json Schema

Evidence base, flat

One row per key finding, with study, DOI, claim, figures, source location and platforms.

evidence.csv Schema

Help Me Combine data

Per-platform and per-pair coverage, quality and cost values used by the pairing engine. Some fields are derived from ChEMBL (CC BY-SA 3.0); see third_party.

helpMeCombine.json Schema

Coverage counts

Distinct proteins overall and per platform, counted from the protein catalogue.

coverage_counts.json Schema

Protein catalogue

Per-protein coverage: gene, UniProt accessions and the platforms that list each protein, compiled from vendor catalogues.

proteins.json Schema

Reactome pathways

Pathway names and members used by the pathway selector.

reactome_pathways.json Schema

Pathway coverage

Precomputed per-pathway coverage for each platform and pair.

pathway_coverage.json Schema

Help Me Choose engine specification

Every scoring constant the Help Me Choose engine uses.

choose_engine_spec.json Schema

Help Me Combine engine specification

Axes, weights and pair values the Help Me Combine engine uses. Some fields are derived from ChEMBL (CC BY-SA 3.0); see third_party.

combine_engine_spec.json Schema

URLs and versions

Files at /data/ are always the latest release. Files at /data/v4.6/ are frozen and do not change, so cite or pin those when you need a result to stay reproducible. The build fails if published data changes without a new version.

The manifest lists every file in the release with its schema, size and SHA-256.

MCP server

aptatlas-mcp lets AI assistants and other tools query this data and run the same recommendation engines as this site. It runs on your own machine from data pinned to one release, makes no network calls and logs nothing. Its results are tested against the published fairness files, byte for byte.

Install and configuration instructions are in its README.

Licence

The files are released under CC BY 4.0, with these sources noted in each file's third_party field:

ChEMBL (CC BY-SA 3.0): some fields in helpMeCombine.json and combine_engine_spec.json are derived from ChEMBL. Each file lists them by path.
UniProt (CC BY 4.0): accessions, gene symbols and protein names.
Reactome (CC0): pathway membership.

The Unassayed Proteome list is not part of this release. Its files and their attribution stay on that page.

Citation

Whelan CD, Smith-Byrne K. Atlas of Proteomic Technologies: an evidence-based framework for selecting and combining commercial proteomics platforms. bioRxiv 2026. https://doi.org/10.64898/2026.09.10.750723

Software and data archive, all versions: https://doi.org/10.5281/zenodo.22697191

Where the scores come from

Claude drafted the first scores, and the authors scored the platforms before opening that draft, then compared the two sets and adjudicated. Every score change, before and since release, was decided by Whelan and Smith-Byrne; Claude made edits only on their instruction. The prompt file generated in the same session as the draft also told the build to adjust scores so that platform totals fell between 28 and 33; the authors did not set this range or use platform totals when scoring or adjudicating.

class_balance, a Help Me Combine axis, is not one of the ten scored dimensions. It is computed, not assigned: the normalised Shannon entropy of the target classes among the drug targets a pair covers, using nine ChEMBL/UniProt classes with Unclassified excluded, divided by ln(9). It is therefore derived from the ChEMBL drug-target list. The per-protein list is not published here, so the values can be checked but not regenerated from these files alone.

The full account is in Methods, and the scripts that regenerate the fairness results are on Reproducibility. In evidence.json, a claim with a source.version of unverified has not been confirmed against the published version of the paper.

Changelog

v4.6, November 4, 2026: First release of the data files and schemas, and of the aptatlas-mcp server.

The site's full version history is in Methods.