遇见数据集

MSA-Readiness Calibration v1 — Method Lock (retrospective) for Protease Structure-Prediction Audit (OpenFold 2.2.0) repository snapshot 2026-09-20, redacted re-publication (2026-09-24)

收藏
Zenodo2026-09-24 更新2026-10-01 收录
官方服务:

资源简介:

Re-publication (2026-09-24): redacted snapshot of the 2026-09-20 update This record republishes the 2026-09-20 whole-repository snapshot. It replaces an earlier upload of the same snapshot that was published as 10.5281/zenodo.22858626 and has since been removed from Zenodo (the record endpoint now returns HTTP 410). The republished archive is not byte-identical to that removed upload: two text members have been redacted, which is the only content change. Everything else, including the member set, is unchanged. What was redacted, and why Two text members recorded the hostname and port of the rented compute node used during the work: - docs/PATH_A_SERVER_BOOTSTRAP.md — 3 host and 3 port occurrences. - reports/path_a_wave1/executed_design.json — 1 host and 1 port occurrence. The values were replaced by the markers [COMPUTE-HOST-REDACTED], [COMPUTE-PORT-REDACTED] and [COMPUTE-CONTAINER-REDACTED]. The source repositories were not modified; redaction was applied to archive member bytes at packaging time only. The pre-redaction SHA-256 and byte count of each redacted member are recorded inside the archive in REDACTION_NOTICE.md and as a separate record file in REDACTION_MAP.tsv, so the change is fully traceable without the redacted values being restated in this description. Consequence for two integrity manifests, registered rather than rewritten. reports/path_a_wave1/manifest.sha256 and reports/path_a_wave1/review/review_package_manifest.json record SHA-256 values for the two redacted members. Those recorded values are the pre-redaction digests and therefore no longer match the packaged bytes. Following this repository's own standing precedent on stale snapshots (docs/复核包清单陈旧哈希登记_2026-09-13.md), the affected manifests were left byte-identical and the discrepancy is registered in REDACTION_NOTICE.md. The JSON manifest additionally carries a machine-readable top-level _c_prime_annotation key naming the superseded digest. Differences relative to the last surviving version (10.5281/zenodo.22723680, v1-postfreeze-2026-09-12) That package has 3,389 members; this package has 3,436 members (206,715,877 bytes compressed; 440,255,617 bytes uncompressed). The member-by-member comparison is distributed with this record as four tables — _A_list.tsv, _B_list.tsv, _C_list.tsv, _D_list.tsv: - A unchanged — 3,358 files: byte-identical (same SHA-256) to that version. - B updated — 31 files: same path, different SHA-256; both digests are in _B_list.tsv. These are the five corrected figure PNGs under reports/path_a_wave1/figures/; the recomputed reports/path_a_wave1/manifest.sha256; the report and review documents (README, review package manifest, review task card, ledger review matrix, number reconciliation, wave1 conclusions draft, wave1 report v2 and v3); the operation logs and the deliverables index; the parameter lock and weight-provenance files; the site curation card; five plotting and verification scripts; the manuscript in both copies; and — newly added to this class by the redaction — docs/PATH_A_SERVER_BOOTSTRAP.md and reports/path_a_wave1/executed_design.json. - C added — 47 files (5,188,897 bytes): present here, absent from that version; listed in _C_list.tsv. They are: (i) the figure vector set re-exported from source — 21 files under reports/path_a_wave1/figures/vector/, being 7 figures × {.svg, .pdf, .dpi300.png}; (ii) 19 correction and audit documents under docs/; (iii) the record lock for the previous version, reports/path_a_wave1/external_freeze/zenodo_22723680_record_lock.json; (iv) the two reviewer-facing errata notices; (v) the journal submission templates; (vi) scripts/rebuild_review_package_manifest.py; and (vii) the package-level REDACTION_NOTICE.md added by this repackaging. - D removed — 0 files: every member of that package is present here. Nature of the corrections carried in this snapshot The repository documents that, after the 2026-09-12 release, several in-figure numeric labels were found to disagree with their source CSVs. Those labels were recomputed from source (or restored where a proposed change was subsequently withdrawn after re-checking), the affected figures and their vector twins were regenerated, the affected reports and review documents were annotated or corrected, and the internal review package was re-signed. Both the corrections and the non-corrections are recorded in the repository's own errata documents. This record distributes that state; it does not adjudicate it. Boundary statement: the 12 frozen v1 method-lock artifacts and docs/MSA_READINESS_SPEC.md were not modified; offline verification against the v1 attachment (10.5281/zenodo.22545061) returns 12/12 ALL PASS (the frozen spec SHA-256 remains 1453a1ff345669fa3ce4bdafcca169bd75ac06be8f94c9e2b0cb2c5143d6aceb). This version adds no prospective_blind count and changes no decision rule or threshold. No raw model inference outputs are part of this release. Verification This record carries nine files: the archive, the per-member hash manifest MANIFEST_sha256.tsv, the four comparison tables _A_list.tsv … _D_list.tsv, the redaction map REDACTION_MAP.tsv, the archive digest file PACKAGE_SHA256.txt, and the extract-and-rehash record package_verification.json. The manifest is a separate file in this record rather than inside the archive, so that it can list every member without a self-reference. Extract the archive and recompute SHA-256 for each member: 3,436 / 3,436 match, with zipfile.testzip() returning clean. Licensing and rights Code is released under the Apache License 2.0; data and text under Creative Commons Attribution 4.0 International (CC-BY-4.0) — the same terms as the previous versions. Copyright (c) 2026 Yao, Jian; Shi, Chenyang; Li, Zhiyi. Upstream third-party data (RCSB PDB, UniProt, PDBe, and others) remain under their own terms and are not re-licensed here. Integrity Archive SHA-256: 9452340b29c9eee5d05fd3126559ad0df06419ed1ff514eec2ef3ac6ac29e03e; MD5: fac0091bd2fd9695afdc1f094c43c83d. How to cite Please cite this version, or the concept DOI 10.5281/zenodo.22545060 for all versions.

提供机构:
Zenodo
创建时间:
2026-09-24
二维码
社区交流群
二维码
科研交流群
商业服务