Peptide Library Release Notes
Jurassic Meat Peptide Library
Internal reference set for phylogenetic reconstruction
Versions 1.0 – 3.2 · 2014–2026
Current release: v3.2 · build 74c9d0b · released 2026-08-18
The peptide library is the versioned reference set used by our Shenzhen Research Center to reconstruct the muscle biology of extinct species. It combines ancient peptide fragments recovered from fossil material with reference sequences from living birds and crocodilians. Library contents are confidential. These release notes are published so that partners, regulators, and other interested parties can follow how the reference set has changed over time. For background, see Science / How It Works.
All prior versions remain archived. Production cell lines reference v3.0 or later.
Each release is accompanied by a technical report in the JM-TR series. Public summaries of those reports are listed in the Technical Reports index.
Contents
- About the library
- Versioning, deprecation and archive policy
- Summary metrics by version
- Known issues (open)
- Deprecations schedule
- Release history (v3.2 → v1.0)
1 · About the library
The library is a single, versioned data product maintained by the Computational Genomics group at the Shenzhen Research Center. It has three components:
- Fossil-derived peptide entries. Ancient peptide fragments identified by LC-MS/MS from demineralised fossil material, each with its spectra-level evidence (peptide-spectrum matches at 1% false discovery rate, or de novo sequence with a stated confidence), its deamidation profile, its procedural blank and contamination screening record, and, since v2.3, its chain-of-custody metadata. The large majority of entries derive from collagen type I (COL1A1 and COL1A2). Tooth-derived entries also include enamel proteins (amelogenin, enamelin, ameloblastin). A small number derive from keratin-associated and non-collagenous bone proteins.
- Extant reference sequences. Protein reference sequences from living birds and crocodilians, the two surviving branches of the archosaur bracket. These carry most of the information in any reconstruction. Fossil peptides constrain it.
- Alignment and placement layer. Since v2.0, a versioned multiple sequence alignment (MSA). Each fossil-derived entry is placed on the reference phylogeny, with a posterior placement probability. Build profiles (theropod, and since v3.1 sauropodomorph) define which entries and constraints are used when the reconstruction pipeline produces a target sequence set.
The library contains peptide and protein sequence data only. It contains no ancient DNA. Ancient DNA is not recoverable from material of this age, and the library has never contained any. Target sequences for cell-line construction are produced computationally from the library by phylogenetic bracketing and ancestral sequence reconstruction. They are not "recovered."
Access is restricted to authorized research staff. Third parties, including partners and regulators, receive build manifests (entry counts, checksums, build profile, commit hash) on request. They do not receive library contents.
2 · Versioning, deprecation and archive policy
2.1 Version numbers
Library releases follow semantic versioning in the form MAJOR.MINOR (with MAJOR.MINOR.PATCH for
patch releases).
- MAJOR is incremented for any breaking change: a change to the alignment or entry format, to the build-profile schema, or to how downstream consumers must reference the library. Breaking changes require a migration note and a technical report.
- MINOR is incremented for additions and non-breaking changes. Examples include new fossil-derived entries, new reference taxa, new metadata fields that are optional for consumers, new build profiles, and changes to placement scoring that do not alter output formats.
- PATCH releases correct metadata, documentation or tooling without changing any sequence, alignment column or placement. Patch releases are recorded in the build log. They are not listed separately in these notes.
Version numbers in these notes are not consecutive. Intermediate minor versions (v1.1–v1.3 and
v2.1–v2.7) were internal maintenance releases. They are archived and are described in the technical
reports, but are not listed separately here; each public entry below summarises all changes since the
previous public entry. This is why the public sequence runs 1.0, 1.4, 2.0, and so on. Release-candidate
numbers (for example v2.0-rc2) are never reused.
2.2 What counts as a breaking change
A change is breaking if a build that previously completed would now fail, or would now produce a different target sequence set from identical inputs without any configuration change by the consumer. A change to placement scoring or to the reference set that moves a posterior, or moves a residue in a re-derived target sequence, is not breaking: production cell lines are pinned to the library state they were constructed or attested against (section 2.5), and every such change is assessed against each production line's acceptance criteria and reported in the technical report. A library release never changes the sequence a production cell line expresses.
2.3 Deprecation policy
Features, formats and flags are deprecated before removal. A deprecated item:
- remains functional for at least 24 months from the release that deprecates it;
- produces a validation warning in every build that uses it;
- is listed in the Deprecations schedule (section 5) with its earliest removal date;
- is removed only in a release that states the removal under Changed or in its migration notes.
The 24-month window may be extended. It is not shortened.
2.4 Archive policy
Every published release is retained in full, read-only, with its alignment, entry set, build profiles, validation report and checksum manifest (SHA-256 per file). Archived releases cannot be modified. A correction to an archived release is made by publishing a new release, never by editing the old one. Archives are held in two physically separate locations in Shenzhen, with a third, offline copy verified annually.
Archived releases remain buildable for audit and reproduction. They are not used for new cell-line construction.
2.5 How production cell lines pin a version
Since v3.0, every active cell-line construction must reference both a library version and a library commit hash (the build ID listed against each release below, in 7-character short form; the full 64-character SHA-256 digest is recorded in the cell-line record). A version string alone is not sufficient, and the construction record is rejected at validation without a hash.
A cell line's library reference is fixed at construction and does not change automatically when a new library release is published. Moving an existing line to a newer library release requires either a new construction record and a new validation cycle, or re-attestation: the target is re-derived from the newer release and compared with the sequence the line expresses, against the line's acceptance criteria (JM-TR-2025-007). Re-attestation does not change the line. Library releases do not change any cell line already in production.
Current production lines reference v3.0 (5d3b8f1) or later. No production line references a release
earlier than v3.0.
3 · Summary metrics by version
Headline figures for each release. Fragment counts and residue coverage refer to theropod-assigned, fossil-derived entries that passed authentication and contamination screening at the time of release. Median placement posterior is calculated over the same set. Reference taxa are extant species contributing reference sequences.
| Version | Released | Build ID | Theropod fragments | Unique residues covered | Median placement posterior (theropod) | Reference taxa (extant) | Status | Technical report |
|---|---|---|---|---|---|---|---|---|
| v1.0 | 2014-03-17 | 3f9a1c2 |
214 | 1,880 | 0.41 | 12 | Archived | JM-TR-2014-002 |
| v1.4 | 2016-11-08 | 8b04e7d |
389 | 3,050 | 0.47 | 19 | Archived | JM-TR-2016-003 |
| v2.0 | 2019-06-04 | c51d2a9 |
1,284 | 7,920 | 0.63 | 34 | Archived | JM-TR-2019-004 |
| v2.3 | 2021-02-22 | 1e7f6b3 |
1,610 | 9,410 | 0.66 | 41 | Archived | JM-TR-2021-005 |
| v2.8 | 2023-09-12 | a92c04e |
2,236 | 12,780 | 0.71 | 57 | Archived | JM-TR-2023-006 |
| v3.0 | 2025-01-20 | 5d3b8f1 |
2,903 | 15,960 | 0.74 | 63 | Supported (production) | JM-TR-2025-007 |
| v3.1 | 2025-10-14 | e06a7c4 |
2,903 (+0 sauropodomorph) | 15,960 | 0.74 | 71 | Supported (production) | JM-TR-2025-008 |
| v3.2 | 2026-08-18 | 74c9d0b |
3,377 | 17,410 | 0.76 | 88 | Current | JM-TR-2026-009 |
The jump between v1.4 and v2.0 reflects two things. First, new fossil material: 383 entries from five specimens. Second, re-search of archived raw data from the four v1.4 specimens under the v2.0 identification workflow: 512 entries. Separately, 141 of the 389 v1.4 entries had been silently dropped by the pre-v2.0 bracketing step (see v2.0, Fixed). They were already counted in v1.4 and are not part of the jump.
Source data
Library version Theropod fragments Unique residues covered v1.0 214 1,880 v1.4 389 3,050 v2.0 1,284 7,920 v2.3 1,610 9,410 v2.8 2,236 12,780 v3.0 2,903 15,960 v3.1 2,903 15,960 v3.2 3,377 17,410
Source data
Library version Median posterior v1.0 0.41 v1.4 0.47 v2.0 0.63 v2.3 0.66 v2.8 0.71 v3.0 0.74 v3.1 0.74 v3.2 0.76
Figure 2 values are as reported at the time of each release. Medians are not directly comparable across the v2.8 boundary, because scoring changed to model diagenetic modification. Re-scoring the v2.3 entry set under the v2.8 model gives a median of 0.68.
4 · Known issues (open)
Issues affecting the current release. Closed issues appear in the release history under Fixed.
| ID | Opened | Since | Description | Impact on production builds | Status |
|---|---|---|---|---|---|
| KI-0412 | 2021-02-22 | v2.3 | Certain keratin-associated peptides produce low-confidence placement. | Entries are flagged and included in default builds at full weight. A profile excluding them is available and is not the default. | Flagged, not removed. |
| KI-0437 | 2021-11-30 | v2.3 | Fragments spanning the COL1A2 telopeptide/helix boundary show placement posteriors sensitive to gap-opening penalty (± 0.09 across the validated penalty range). | None identified. Boundary columns are masked in the theropod build profile. | Under investigation. |
| KI-0519 | 2023-09-12 | v2.8 | Racemization (Asx D/L) data are unavailable for 7 of 21 specimens, including the founding specimen, which was not subsampled for chiral analysis. The diagenesis model estimates the diagenetic index for these specimens without chiral data. | None identified. | Accepted. |
| KI-0588 | 2025-10-14 | v3.1 | Sauropodomorph build profile reports "N/A" for median fossil placement posterior, because it contains no fossil-derived sauropodomorph entries. | Expected behaviour. See v3.1. | Accepted. |
| KI-0601 | 2026-08-18 | v3.2 | Very short peptide fragments (<8 residues) remain high-variance in placement. | Short fragments carry reduced weight in bracketing. See JM-TR-2026-009. | Documented. Not scheduled for removal. |
| KI-0604 | 2026-08-18 | v3.2 | Two paleognath reference taxa added in v3.2 have partial COL1A2 references (≥ 92% length). Coverage is flagged in the manifest. | None identified. | Open. |
5 · Deprecations schedule
| Item | Deprecated in | Earliest removal | Removed in | Notes |
|---|---|---|---|---|
Unaligned legacy peptide files (.jmpep, flat FASTA with free-text headers) |
v2.0 (2019-06) | 2021-06 | v2.8 (2023-09) | Converted to MSA entries by jmlib-migrate. Originals retained in the v1.x archives. |
--no-bracket-filter build option |
v2.0 (2019-06) | 2021-06 | v2.3 (2021-02, disabled); v2.8 (removed) | Disabled early, outside the 24-month window, under the v2.0 bracketing fix. The only exception to the deprecation window to date. |
Free-text specimen provenance field (origin_note) |
v2.3 (2021-02) | 2023-02 | v3.0 (2025-01) | Replaced by structured chain-of-custody fields. Free-text content preserved as origin_note_legacy, read-only. |
| Manual override flags older than 24 months | v2.8 (2023-09) | 2025-09 | v3.1 (2025-10) | Overrides must now be renewed with a named reviewer and a reason. Expired overrides are ignored at build time. |
Version-only library references in cell-line records (e.g. library: "2.8") |
v3.0 (2025-01) | 2027-01 | Not yet removed | Still accepted for archived and research-only records, with a validation warning. Rejected for active constructions since v3.0. |
Implicit theropod build profile (builds with no profile: key) |
v3.1 (2025-10) | 2027-10 | Not yet removed | Default profile will be removed. Builds must name a profile explicitly. |
6 · Release history
v3.2 (2026-08)
Released: 2026-08-18 · Build: 74c9d0b · Type: minor
Compatibility: Non-breaking. Cell-line constructions pinned to v3.0 or v3.1 are unaffected. No
changes to downstream cell-line validation criteria.
Technical report: JM-TR-2026-009, paleognath reference
expansion and short-fragment placement variance.
Added
- Minor: updated avian reference set to include additional paleognath sequences. Reference taxa increase from 71 to 88, principally ratites and tinamous.
- 474 theropod-assigned fossil-derived entries (2,903 → 3,377), extending residue coverage from 15,960 to 17,410.
- Per-entry fragment-length field in the build manifest, to support filtering of short fragments by consumers.
- Reference-completeness flag for extant reference sequences below full length.
Changed
- Short fragments (<8 residues) now carry reduced weight in bracketing, according to the per-entry placement variance. The production line DN-2019-04 remains within its acceptance criteria and continues to reference v3.0.
- Validation report now lists placement posterior distributions by fragment-length class.
Known issues
- Known issue: very short peptide fragments (<8 residues) remain high-variance in placement. Documented. Not scheduled for removal. (KI-0601)
- Two newly added paleognath taxa have partial COL1A2 references. (KI-0604)
Notes
- No changes to downstream cell-line validation criteria.
v3.1 (2025-10)
Released: 2025-10-14 · Build: e06a7c4 · Type: minor
Compatibility: Non-breaking. Theropod build outputs are identical to v3.0 for identical
configuration (verified by checksum across all production profiles).
Technical report: JM-TR-2025-008, the sauropodomorph
build profile and reference-only bracketing constraints.
Added
- Added a sauropodomorph-targeted build profile to support the Brachiosaurus program. No sauropodomorph-derived peptides are included in this release; the profile currently relies on reference sequences and bracketing constraints only.
- Eight extant reference taxa (63 → 71), selected to improve bracketing depth for the sauropodomorph profile.
profile:key in build configuration. Builds may now name a profile explicitly.
Changed
- Manual override flags older than 24 months (deprecated in v2.8) are now ignored at build time.
- Validation report states, for each profile, the number of fossil-derived entries contributing to the target sequence set. For the sauropodomorph profile this number is zero.
Deprecated
- Builds with no
profile:key currently default to the theropod profile. This implicit default is deprecated. Earliest removal: 2027-10.
Known issues
- Sauropodomorph profile reports "N/A" for median fossil placement posterior. (KI-0588)
Documentation
- Documentation updates to build-filter configuration.
- Added worked examples of profile selection, including an example showing a mis-set filter rejected at validation.
v3.0 (2025-01)
Released: 2025-01-20 · Build: 5d3b8f1 · Type: major (breaking)
Compatibility: Breaking for cell-line construction records. The library entry and alignment
formats are unchanged from v2.8.
Technical report: JM-TR-2025-007, commit-hash provenance
for cell-line constructions and the cross-clade build-filter defect.
Breaking
- Breaking change
- all active cell-line constructions are now required to reference a library commit hash.
Added
- 667 theropod-assigned fossil-derived entries (2,236 → 2,903). Six extant reference taxa (57 → 63).
- Clade-assertion check at build time. A build fails if any contributing entry's placement lies outside the clade named by the build profile.
- Signed checksum manifest for each release (SHA-256 per file, manifest signed by the library maintainers' release key).
Fixed
- Fixed an issue where non-theropod peptides could be included in theropod-targeted builds if filters were mis-set. Scope, root cause and corrective action are described in JM-TR-2025-007. All production constructions were re-verified against the corrected filter before this release.
- Fixed a timezone handling error that recorded some build timestamps as the following day. Archived build logs were annotated, not altered.
Changed
- Free-text
origin_note(deprecated in v2.3) removed. Content preserved asorigin_note_legacy.
Deprecated
- Version-only library references in cell-line records. Earliest removal: 2027-01.
Security & integrity
- Write access to the release branch reduced to two named maintainers. A release now requires approval from a maintainer who did not prepare it.
- Build hosts no longer have outbound network access during builds.
Migration notes
- Every active cell-line construction record must add
library_commit(full 64-character SHA-256 digest) alongsidelibrary_version. Records without it fail validation. - For records created against v2.8 as released, the hash is
a92c04e…. The label v2.8 also referred to six later content states (JM-TR-2025-007, Table 1);jmlib-pinselects among them by build date or manifest checksum. Full digests are in the v2.8 archive manifest.jmlib-pinpopulates the field from an existing version string and records who ran it. - Records for archived or research-only lines may keep version-only references until the deprecation window closes, with a validation warning.
v2.8 (2023-09)
Released: 2023-09-12 · Build: a92c04e · Type: minor
Compatibility: Non-breaking. Placement posteriors change under the new scoring model. Re-derived
theropod target sequences changed at 12 positions in five targets; impact assessment IA-2023-03 found no
impact on released material.
Technical report: JM-TR-2023-006, modelling diagenetic
modification in placement scoring.
Added
- 626 theropod-assigned fossil-derived entries (1,610 → 2,236). Sixteen extant reference taxa (41 → 57).
- Per-entry diagenesis annotations: deamidation (N/Q), oxidation, backbone cleavage sites, and, where measured, racemization.
Changed
- Improved handling of diagenetically altered residues. Placement scoring now models deamidation, oxidation and backbone cleavage explicitly, rather than treating altered residues as mismatches.
- Unaligned legacy peptide files (deprecated in v2.0) removed. All legacy content had been migrated.
Fixed
- Contamination exclusion list was not applied to identifications supported by de novo sequencing alone. Applied retrospectively: affected entries were re-screened and the reclassified entries are listed in the v2.8 validation report. No affected entry contributed to a production target sequence.
- Build manifest reported reference taxa counts that included two duplicate species entered under a former and a current scientific name. Corrected.
Deprecated
- Deprecated manual override flags older than 24 months. Earliest removal: 2025-09. (Removed in v3.1.)
Known issues
- Racemization estimates unavailable for seven specimens, including the founding specimen. (KI-0519)
v2.3 (2021-02)
Released: 2021-02-22 · Build: 1e7f6b3 · Type: minor
Compatibility: Non-breaking. New metadata fields are optional for consumers and mandatory for new
fossil-derived entries.
Technical report: JM-TR-2021-005, low-confidence
keratin-associated placement and the chain-of-custody metadata schema.
Added
- Added chain-of-custody metadata fields for all fossil-derived entries. Fields include specimen identifier, acquisition date, acquisition source, documented locality (where available), transfer history, storage conditions and sampling record.
- 326 theropod-assigned fossil-derived entries (1,284 → 1,610). Seven extant reference taxa (34 → 41).
- Low-confidence flag (
placement_flag, valueLOWCONF_KAP) on entries, visible in manifests and validation reports.
Changed
--no-bracket-filterdisabled ahead of its scheduled removal (see Deprecations schedule).
Fixed
- Specimen identifiers for two v1.4 entries had been transposed. Corrected. Placement is unaffected, because both entries derive from the same specimen batch.
Deprecated
- Free-text
origin_noteprovenance field, replaced by structured chain-of-custody fields. Earliest removal: 2023-02. (Removed in v3.0.)
Known issues
- Known issue: certain keratin-associated peptides produce low-confidence placement. Flagged, not removed. (KI-0412)
- COL1A2 telopeptide/helix boundary placements are sensitive to gap-opening penalty. (KI-0437, opened 2021-11-30)
Notes
- For some entries, chain-of-custody fields could only be partially completed from existing records. For specimens acquired before 2016, acquisition source is recorded at the level of documentation available at the time.
v2.0 (2019-06)
Released: 2019-06-04 · Build: c51d2a9 · Type: major (breaking)
Compatibility: Breaking. The library format changes from per-specimen peptide files to a
versioned multiple sequence alignment. v1.x consumers must migrate.
Technical report: JM-TR-2019-004, the versioned MSA
architecture and post-mortem of the fragment-dropping defect.
Breaking
- Breaking change
- migrated to a versioned multiple-sequence alignment format.
Added
- 895 theropod-assigned fossil-derived entries (389 → 1,284): 383 from five new specimens and 512 from re-search of archived raw data. Fifteen extant reference taxa (19 → 34).
- Phylogenetic placement with posterior probabilities for every fossil-derived entry, replacing curator acceptance of computed placements.
- Automated validation report for every build.
jmlibcommand-line tooling for builds, manifests and diffs between releases.
Fixed
- Fixed an issue where fragmentary sequences were silently dropped during bracketing. Affected releases v1.1–v1.4 (141 of 389 entries in the v1.4 build); in service August 2014 to November 2018. Impact assessment: no impact on released material. Full post-mortem in JM-TR-2019-004.
Deprecated
- Deprecated support for unaligned legacy peptide files. Earliest removal: 2021-06. (Removed in v2.8.)
--no-bracket-filterbuild option.
Migration notes
- Legacy peptide files are converted with
jmlib-migrate, which writes a per-file report of entries aligned, entries flagged for curator review, and entries not aligned. Entries not aligned are listed, never discarded. - Builds referencing v1.x continue to run against the v1.x archives for audit purposes only.
v1.4 (2016-11)
Released: 2016-11-08 · Build: 8b04e7d · Type: minor
Compatibility: Non-breaking. Same file formats as v1.0.
Technical report: JM-TR-2016-003, fragments from three
additional specimens and specimen-acquisition practice.
Added
- Added first fragments from additional specimens, acquired with documented provenance. Low coverage. Three specimens; 88 theropod-assigned entries. With 87 entries from further extraction of the founding specimen in internal releases v1.1–v1.3, the library grows from 214 to 389 entries.
- Seven extant reference taxa (12 → 19).
- Procedural blank results recorded per extraction batch, alongside each entry.
Changed
- Automated bracketing module in use for builds since v1.1, replacing manual entry of constraints.
- Curator disagreements on placement are now recorded alongside the entry.
Fixed
- Human keratin and porcine trypsin contamination databases were updated to current versions. The previous versions dated from 2012.
Notes
- Placement in this release was computed after curation and accepted or overridden by curators. Posterior probabilities shown in Table 1 for v1.0 and v1.4 were computed at release by the v1.x placement pipeline and are retained for comparison.
v1.0 (2014-03)
Released: 2014-03-17 · Build: 3f9a1c2 · Type: initial release
Compatibility: First structured release. No earlier versioned release exists.
Technical report: JM-TR-2014-002, the first structured
library. Background: JM-TR-2013-001, on the founding
recovery and authentication of peptide fragments from a Kem Kem Group theropod tooth.
Added
- First structured release. Consolidates the tooth-derived theropod peptide fragments recovered in 2011–2012, previously maintained as unversioned working files.
- Initial import of avian and crocodilian reference sequences. Twelve extant reference taxa.
- 214 theropod-assigned fossil-derived entries covering 1,880 unique residues, principally from COL1A1, COL1A2 and enamel proteins.
- Deamidation (N/Q) ratio recorded per entry as an authenticity marker.
Notes
- Manual curation only.
- Working files from 2011–2012 were retained unmodified and archived alongside this release.
Questions about this document may be directed to regulatory@jurassicmeat.com. Library access is restricted to authorized research staff and is not available to third parties.
Bringing the Taste of the Past, Tomorrow.