Evidence claim

ProtGPT2: description

Description of ProtGPT2

Deep unsupervised language model trained on protein sequences that generates de novo protein sequences, scored in ProteinGym as a zero-shot single-sequence substitution baseline.

Replace a placeholder or provenance description with a sourced functional description (issue #88a). Historical releases retain the original value.

Evidence

Source checking verifies the cited claim or transcription. It does not establish independent reproduction.

Evidence table

Inspect claims, sources and review details

Trace each statement to its source and review. A context-only reference supports the record generally; it does not verify an individual field. Source checking does not reproduce an experiment.

One row per statement and cited source. Multiple citations are not independent evaluations. Shared locators are labelled explicitly.

21 evidence rows matching the loaded filters

Claims, original sources and review scope · Release 2026-10-22-1b7cf027e4f6
Property and statement	Original source and location	Review and provenance
Claimed field description Context-only references	Abstract: ProtGPT2 is a deep unsupervised language model for protein design Original source ↗ uc20260930-source-proteingym-amfr-spearman: Benchmark CSV header (model scored in the zero-shot DMS substitution benchmark); run-doc-proteingym-readme-md-144fe22b: README 'Leaderboard' baseline table (model name, input modalities, reference); qa-descriptions-20261010-source-protgpt2: abstract Shared locator for this statement’s cited sources; not a separate locator for each citation. Version: Europe PMC core record MED/35896542 (title and abstract), Nature communications 2022 Retrieved: 2026-10-10T06:44:11Z	not individually reviewed No individual claim review recorded Audit details Field: `attributes.field` Source artifact SHA-256: `90fccf89e2600cc576108d6a6ded86f9577c4a20dae8a60b06b35c7ba8dc287f` Hash scope: SHA-256 of one retrieval of a Europe PMC REST search response. The response is not byte-stable: on 2026-10-10 a second retrieval of 10 of these 31 records differed only in JSON key order and parsed to identical content. The title and abstractText fields are the evidence. Inspected artifact
Claimed field description Context-only references	OATML-Markslab/ProteinGym / README.md Original source ↗ uc20260930-source-proteingym-amfr-spearman: Benchmark CSV header (model scored in the zero-shot DMS substitution benchmark); run-doc-proteingym-readme-md-144fe22b: README 'Leaderboard' baseline table (model name, input modalities, reference); qa-descriptions-20261010-source-protgpt2: abstract Shared locator for this statement’s cited sources; not a separate locator for each citation. Version: Git commit 144fe22b07dfaeec2b366f2346203a9838a55b4c Retrieved: 2026-09-17T09:32:03.517591+00:00	not individually reviewed No individual claim review recorded Audit details Field: `attributes.field` Source artifact SHA-256: `321487a8de52c6cfa647a658f61150dd72e0acb0125470524fb94d1f8b23321a` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Claimed field description Context-only references	ProteinGym AMFR assay-level Spearman comparison Original source ↗ uc20260930-source-proteingym-amfr-spearman: Benchmark CSV header (model scored in the zero-shot DMS substitution benchmark); run-doc-proteingym-readme-md-144fe22b: README 'Leaderboard' baseline table (model name, input modalities, reference); qa-descriptions-20261010-source-protgpt2: abstract Shared locator for this statement’s cited sources; not a separate locator for each citation. Version: 144fe22b07dfaeec2b366f2346203a9838a55b4c Retrieved: 2026-09-30	not individually reviewed No individual claim review recorded Audit details Field: `attributes.field` Source artifact SHA-256: `f432423b87f79ac9778dfac86e3d95be041d246bc618dc0a406b35b0b7466437` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Previous value Named prediction model family identified in the cited primary source. Exact fitted configurations and scores are separate records. Context-only references	Abstract: ProtGPT2 is a deep unsupervised language model for protein design Original source ↗ uc20260930-source-proteingym-amfr-spearman: Benchmark CSV header (model scored in the zero-shot DMS substitution benchmark); run-doc-proteingym-readme-md-144fe22b: README 'Leaderboard' baseline table (model name, input modalities, reference); qa-descriptions-20261010-source-protgpt2: abstract Shared locator for this statement’s cited sources; not a separate locator for each citation. Version: Europe PMC core record MED/35896542 (title and abstract), Nature communications 2022 Retrieved: 2026-10-10T06:44:11Z	not individually reviewed No individual claim review recorded Audit details Field: `attributes.previous_value` Source artifact SHA-256: `90fccf89e2600cc576108d6a6ded86f9577c4a20dae8a60b06b35c7ba8dc287f` Hash scope: SHA-256 of one retrieval of a Europe PMC REST search response. The response is not byte-stable: on 2026-10-10 a second retrieval of 10 of these 31 records differed only in JSON key order and parsed to identical content. The title and abstractText fields are the evidence. Inspected artifact
Previous value Named prediction model family identified in the cited primary source. Exact fitted configurations and scores are separate records. Context-only references	OATML-Markslab/ProteinGym / README.md Original source ↗ uc20260930-source-proteingym-amfr-spearman: Benchmark CSV header (model scored in the zero-shot DMS substitution benchmark); run-doc-proteingym-readme-md-144fe22b: README 'Leaderboard' baseline table (model name, input modalities, reference); qa-descriptions-20261010-source-protgpt2: abstract Shared locator for this statement’s cited sources; not a separate locator for each citation. Version: Git commit 144fe22b07dfaeec2b366f2346203a9838a55b4c Retrieved: 2026-09-17T09:32:03.517591+00:00	not individually reviewed No individual claim review recorded Audit details Field: `attributes.previous_value` Source artifact SHA-256: `321487a8de52c6cfa647a658f61150dd72e0acb0125470524fb94d1f8b23321a` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Previous value Named prediction model family identified in the cited primary source. Exact fitted configurations and scores are separate records. Context-only references	ProteinGym AMFR assay-level Spearman comparison Original source ↗ uc20260930-source-proteingym-amfr-spearman: Benchmark CSV header (model scored in the zero-shot DMS substitution benchmark); run-doc-proteingym-readme-md-144fe22b: README 'Leaderboard' baseline table (model name, input modalities, reference); qa-descriptions-20261010-source-protgpt2: abstract Shared locator for this statement’s cited sources; not a separate locator for each citation. Version: 144fe22b07dfaeec2b366f2346203a9838a55b4c Retrieved: 2026-09-30	not individually reviewed No individual claim review recorded Audit details Field: `attributes.previous_value` Source artifact SHA-256: `f432423b87f79ac9778dfac86e3d95be041d246bc618dc0a406b35b0b7466437` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Source location uc20260930-source-proteingym-amfr-spearman: Benchmark CSV header (model scored in the zero-shot DMS substitution benchmark); run-doc-proteingym-readme-md-144fe22b: README 'Leaderboard' baseline table (model name, input modalities, reference); qa-descriptions-20261010-source-protgpt2: abstract Context-only references	Abstract: ProtGPT2 is a deep unsupervised language model for protein design Original source ↗ uc20260930-source-proteingym-amfr-spearman: Benchmark CSV header (model scored in the zero-shot DMS substitution benchmark); run-doc-proteingym-readme-md-144fe22b: README 'Leaderboard' baseline table (model name, input modalities, reference); qa-descriptions-20261010-source-protgpt2: abstract Shared locator for this statement’s cited sources; not a separate locator for each citation. Version: Europe PMC core record MED/35896542 (title and abstract), Nature communications 2022 Retrieved: 2026-10-10T06:44:11Z	not individually reviewed No individual claim review recorded Audit details Field: `attributes.source_locator` Source artifact SHA-256: `90fccf89e2600cc576108d6a6ded86f9577c4a20dae8a60b06b35c7ba8dc287f` Hash scope: SHA-256 of one retrieval of a Europe PMC REST search response. The response is not byte-stable: on 2026-10-10 a second retrieval of 10 of these 31 records differed only in JSON key order and parsed to identical content. The title and abstractText fields are the evidence. Inspected artifact
Source location uc20260930-source-proteingym-amfr-spearman: Benchmark CSV header (model scored in the zero-shot DMS substitution benchmark); run-doc-proteingym-readme-md-144fe22b: README 'Leaderboard' baseline table (model name, input modalities, reference); qa-descriptions-20261010-source-protgpt2: abstract Context-only references	OATML-Markslab/ProteinGym / README.md Original source ↗ uc20260930-source-proteingym-amfr-spearman: Benchmark CSV header (model scored in the zero-shot DMS substitution benchmark); run-doc-proteingym-readme-md-144fe22b: README 'Leaderboard' baseline table (model name, input modalities, reference); qa-descriptions-20261010-source-protgpt2: abstract Shared locator for this statement’s cited sources; not a separate locator for each citation. Version: Git commit 144fe22b07dfaeec2b366f2346203a9838a55b4c Retrieved: 2026-09-17T09:32:03.517591+00:00	not individually reviewed No individual claim review recorded Audit details Field: `attributes.source_locator` Source artifact SHA-256: `321487a8de52c6cfa647a658f61150dd72e0acb0125470524fb94d1f8b23321a` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Source location uc20260930-source-proteingym-amfr-spearman: Benchmark CSV header (model scored in the zero-shot DMS substitution benchmark); run-doc-proteingym-readme-md-144fe22b: README 'Leaderboard' baseline table (model name, input modalities, reference); qa-descriptions-20261010-source-protgpt2: abstract Context-only references	ProteinGym AMFR assay-level Spearman comparison Original source ↗ uc20260930-source-proteingym-amfr-spearman: Benchmark CSV header (model scored in the zero-shot DMS substitution benchmark); run-doc-proteingym-readme-md-144fe22b: README 'Leaderboard' baseline table (model name, input modalities, reference); qa-descriptions-20261010-source-protgpt2: abstract Shared locator for this statement’s cited sources; not a separate locator for each citation. Version: 144fe22b07dfaeec2b366f2346203a9838a55b4c Retrieved: 2026-09-30	not individually reviewed No individual claim review recorded Audit details Field: `attributes.source_locator` Source artifact SHA-256: `f432423b87f79ac9778dfac86e3d95be041d246bc618dc0a406b35b0b7466437` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Claimed value Deep unsupervised language model trained on protein sequences that generates de novo protein sequences, scored in ProteinGym as a zero-shot single-sequence substitution baseline. Context-only references	Abstract: ProtGPT2 is a deep unsupervised language model for protein design Original source ↗ uc20260930-source-proteingym-amfr-spearman: Benchmark CSV header (model scored in the zero-shot DMS substitution benchmark); run-doc-proteingym-readme-md-144fe22b: README 'Leaderboard' baseline table (model name, input modalities, reference); qa-descriptions-20261010-source-protgpt2: abstract Shared locator for this statement’s cited sources; not a separate locator for each citation. Version: Europe PMC core record MED/35896542 (title and abstract), Nature communications 2022 Retrieved: 2026-10-10T06:44:11Z	not individually reviewed No individual claim review recorded Audit details Field: `attributes.value` Source artifact SHA-256: `90fccf89e2600cc576108d6a6ded86f9577c4a20dae8a60b06b35c7ba8dc287f` Hash scope: SHA-256 of one retrieval of a Europe PMC REST search response. The response is not byte-stable: on 2026-10-10 a second retrieval of 10 of these 31 records differed only in JSON key order and parsed to identical content. The title and abstractText fields are the evidence. Inspected artifact

How to read the evidence

Sources and history

Release 2026-10-22-1b7cf027e4f6 · Record review: source checked

3 source records and release history

ProteinGym AMFR assay-level Spearman comparison · Original source · 144fe22b07dfaeec2b366f2346203a9838a55b4c
OATML-Markslab/ProteinGym / README.md · Original source · Git commit 144fe22b07dfaeec2b366f2346203a9838a55b4c
Abstract: ProtGPT2 is a deep unsupervised language model for protein design · Original source · Europe PMC core record MED/35896542 (title and abstract), Nature communications 2022

Download this release (gzip)

Technical metadata and extraction receipts

Stable ID: metadata-correction-46d0b40f1618a5ccbf55

areas: proteins-complexes
field: description
previous value: Named prediction model family identified in the cited primary source. Exact fitted configurations and scores are separate records.
value: Deep unsupervised language model trained on protein sequences that generates de novo protein sequences, scored in ProteinGym as a zero-shot single-sequence substitution baseline.
source locator: uc20260930-source-proteingym-amfr-spearman: Benchmark CSV header (model scored in the zero-shot DMS substitution benchmark); run-doc-proteingym-readme-md-144fe22b: README 'Leaderboard' baseline table (model name, input modalities, reference); qa-descriptions-20261010-source-protgpt2: abstract
review: method: automated-source-review; ai-assisted-source-review; reviewer: claude; reviewer note: Extractor: Claude research agent for issue #88a. Reviewer: separate Claude review agent, independent of the extractor; no human review claimed; date: 2026-10-10; method note: Compared the description with the cited source text (abstract, README row, benchmark CSV header or full-text section) of re-downloaded artifacts, and previous_value with origin/main.; note: Description written from the cited source text; provenance text was generic and is kept here as previous_value. Independent review 2026-10-10: corrected. 'Transformer language model' changed to the abstract's 'deep unsupervised language model'; the abstract cites Transformer progress as motivation but does not state ProtGPT2's architecture.

Related records

subject: ProtGPT2

Suggest a correction