申请配额 联系我们

科学技术报告

JM-TR-2026-009 技术报告 文库 v3.2

Library v3.2: palaeognath reference expansion and the placement variance of short peptide fragments

Dr. Elena Popescu 1,* Dr. Anders Kjeldsen 1 Dr. Hao Lin 1

1 侏罗纪肉业 深圳研究中心
* 通讯作者: Dr. Elena Popescu · regulatory@jurassicmeat.com

目录
  1. 摘要
  2. 1. Summary
  3. 2. Introduction
  4. 3. Materials and Methods
  5. 4. Results
  6. 5. Discussion
  7. 6. Limitations and known issues
  8. 7. Changes in Library v3.2
  9. 8. Data availability
  10. 9. Acknowledgements
  11. 10. References
  12. Supplementary Table S1

摘要

Library v3.2 expands the palaeognath component of the avian reference set from five taxa to twenty-two, adding representatives of ostriches, rheas, cassowaries, kiwi and tinamous (17 taxa; 88 extant reference taxa in total), and adds 474 theropod-assigned fossil-derived fragments from two new specimens and from re-extraction of archived material (3,377 fragments; 17,410 unique residues). The median theropod placement posterior rises from 0.74 to 0.76. The improvement is not uniform. For fragments of 8 residues or more, placement posteriors increased in 25 of 26 sampled fragments and replicate variance was unchanged. For fragments shorter than 8 residues, which make up 21.8% of the v3.2 theropod set, median replicate standard deviation of the placement posterior increased from 0.16 to 0.20, and the rate at which replicate placements disagreed with the consensus clade rose from 0.22 to 0.31. We attribute this to the low information content of short collagen fragments, compounded by the dilution of posterior mass across closely related new reference taxa. We evaluate removal of short fragments at thresholds from 6 to 10 residues. At 8 residues, removal would leave 412 currently anchored residue positions without any fossil anchor and would remove 41% of the fragments recovered from the founding specimen. Short fragments are retained with their variance documented. This known issue is documented and is not scheduled for removal.

关键词Palaeognathaephylogenetic placementshort peptidesplacement variancereference taxon samplingtype I collagen

JM-TR-2026-009 · Library v3.2 · 18 August 2026 Dr. Elena Popescu, Dr. Anders Kjeldsen, Dr. Hao Lin Computational Genomics and Palaeoproteomics, Jurassic Meat Shenzhen Research Center

Public summary of internal report (37 pp.). Library contents are confidential.

1. Summary

  • Reference expansion. 17 palaeognath taxa were added to the avian reference set (88 extant reference taxa in total). Each of the five extant palaeognath lineages is now represented by more than one taxon.
  • Fragment growth. 474 theropod-assigned fragments were added (3,377 total; 17,410 unique residues). 211 of the new fragments (44.5%) are shorter than 8 residues.
  • Placement. The median theropod placement posterior rose from 0.74 to 0.76. Fragments of 8 residues or more improved. Fragments shorter than 8 residues became more variable across replicate placements.
  • Known issue retained. Short fragments (<8 residues) remain high-variance in placement. Removal was evaluated and not adopted. Documented. Not scheduled for removal.
  • Downstream. No change to cell-line validation criteria. The production line DN-2019-04 remains within acceptance criteria against v3.2.

2. Introduction

Crown birds divide at their base into Palaeognathae and Neognathae. Palaeognaths (the ostriches, rheas, cassowaries and emu, kiwi, and the volant tinamous) are the earlier-diverging lineage, and their sequences therefore carry disproportionate weight in estimating the ancestral state of crown birds, which is the node through which the avian side of the bracket reaches every non-avian dinosaur target (Witmer, 1995). Until this release the library carried five palaeognaths, one from each extant lineage (Struthio camelus, Rhea americana, Dromaius novaehollandiae, Apteryx australis and Tinamus guttatus). Struthio has been in the reference set since the founding work and Tinamus since v1.3; the other three were added at Library v2.0, where the palaeognaths accounted for much of that release's gain in placement confidence (JM-TR-2019-004). The sensitivity analyses reported for the sauropodomorph profile found that removing them changed only 15 positions in the Brachiosaurus target (JM-TR-2025-008). We read this not as evidence that palaeognaths are unimportant but as evidence that one representative per lineage was too few to matter, and v3.2 was planned accordingly.

The same release adds 474 theropod-assigned fragments, the first addition of fossil-derived fragments since v3.0. The majority come from two specimens acquired in 2025 and from re-extraction of archived specimens using the revised demineralisation procedure introduced in 2024. Revised demineralisation recovers more peptide material from previously extracted specimens, and a disproportionate share of what it recovers is short.

Short fragments have been a known weak point of the library since placement scoring was first formalised. The 8-residue threshold in the original v2.0 placement analysis marks the length below which median posterior placement probability falls under 0.5 (JM-TR-2019-004), and diagenetic modification affects short fragments disproportionately because a single deamidated residue is a larger fraction of the evidence (JM-TR-2023-006). Short-fragment variance has been tracked in the internal known-issue register since v2.0 (KI-2019-02; published as KI-0601 from v3.2). This report quantifies how the v3.2 changes interact with that known issue and records the decision not to remove them.

Short-peptide phylogenetics is not unique to this program. Enamel proteomes from Pleistocene and Miocene material have resolved deep relationships from limited peptide coverage (Cappellini et al., 2019; Welker et al., 2019). Those studies benefited from longer contiguous coverage and younger material than we typically recover from Cretaceous specimens, and their results should not be taken as an indication of what is achievable here.

3. Materials and Methods

3.1 Reference taxa

Seventeen palaeognath taxa were added (Table 1). Reference sequences for the reconstruction targets and for the collagen and enamel proteins against which fossil fragments are placed were assembled from public sequence resources, translated, aligned to the v3.1 alignment by profile alignment, and reviewed manually at every position where the new sequence introduced a gap. Taxa were accepted if they covered at least 90% of target positions. One candidate, a second rhea sample, was replaced after it failed this criterion.

Table 1. Palaeognath reference taxa in Library v3.2.
Lineage Present before v3.2 Added in v3.2 Target-set coverage (median of added taxa)
Struthioniformes (ostriches) Struthio camelus Struthio molybdophanes 98.7%
Rheiformes (rheas) Rhea americana Rhea pennata 96.1%
Casuariiformes (emu, cassowaries) Dromaius novaehollandiae Casuarius casuarius, Casuarius bennetti, Casuarius unappendiculatus 95.4%
Apterygiformes (kiwi) Apteryx australis Apteryx mantelli, Apteryx owenii, Apteryx haastii, Apteryx rowi 94.8%
Tinamiformes (tinamous) Tinamus guttatus Tinamus major, Crypturellus cinnamomeus, Crypturellus undulatus, Crypturellus tataupa, Eudromia elegans, Nothoprocta perdicaria, Nothoprocta pentlandii, Nothura maculosa 93.2%
Total 5 17

The palaeognath topology used is shown in Fig. 1. The arrangement of rheas, tinamous and the emu–cassowary–kiwi clade relative to one another is not settled, and sensitivity to it is reported in Section 4.5.

Figure 1. Palaeognath reference topology in Library v3.2Reduced cladogram (genus level) of the avian reference set near the base of crown birds, with the non-avian theropod placement branch (JM theropod) and Alligator for orientation. Genera new to the library in v3.2 are highlighted; species were also added within Struthio, Rhea, Apteryx and Tinamus. Placement probabilities are shown for a representative 7-residue COL1A2 fragment from the founding specimen (F-01-0117) under v3.2; the remaining posterior mass (0.48) lies on internal branches.Struthiop = 0.14Struthio — placement posterior 0.14Rheap = 0.07Rhea — placement posterior 0.07TinamusTinamusCrypturellusCrypturellusEudromiaEudromiaNothoproctaNothoproctaNothuraNothuraApteryxApteryxDromaiusDromaiusCasuariusCasuariusGallusp = 0.09Gallus — placement posterior 0.09AnasAnasJM theropodp = 0.18JM theropod — placement posterior 0.18Alligatorp = 0.04Alligator — placement posterior 0.04Figure 1. Palaeognath reference topology in Library v3.2Reduced cladogram (genus level) of the avian reference set near the base of crown birds, with the non-avian theropod placement branch (JM theropod) and Alligator for orientation. Genera new to the library in v3.2 are highlighted; species were also added within Struthio, Rhea, Apteryx and Tinamus. Placement probabilities are shown for a representative 7-residue COL1A2 fragment from the founding specimen (F-01-0117) under v3.2; the remaining posterior mass (0.48) lies on internal branches.Struthiop = 0.14Struthio — placement posterior 0.14Rheap = 0.07Rhea — placement posterior 0.07TinamusTinamusCrypturellusCrypturellusEudromiaEudromiaNothoproctaNothoproctaNothuraNothuraApteryxApteryxDromaiusDromaiusCasuariusCasuariusGallusp = 0.09Gallus — placement posterior 0.09AnasAnasJM theropodp = 0.18JM theropod — placement posterior 0.18Alligatorp = 0.04Alligator — placement posterior 0.04
Figure 1.Palaeognath reference topology in Library v3.2. Reduced cladogram (genus level) of the avian reference set near the base of crown birds, with the non-avian theropod placement branch (JM theropod) and Alligator for orientation. Genera new to the library in v3.2 are highlighted; species were also added within Struthio, Rhea, Apteryx and Tinamus. Placement probabilities are shown for a representative 7-residue COL1A2 fragment from the founding specimen (F-01-0117) under v3.2; the remaining posterior mass (0.48) lies on internal branches.Source: Library v3.2, reference alignment and placement run PL-2026-07.
Source data
TaxonHighlightedPlacement posterior
Struthio0.14
Rhea0.07
Tinamus
Crypturellusyes
Eudromiayes
Nothoproctayes
Nothurayes
Apteryx
Dromaius
Casuariusyes
Gallus0.09
Anas
JM theropod0.18
Alligator0.04

((((Struthio,(Rhea,(((Tinamus,Crypturellus),(Eudromia,(Nothoprocta,Nothura))),(Apteryx,(Dromaius,Casuarius))))),(Gallus,Anas)),JM_theropod),Alligator);

3.2 New fossil-derived fragments

Fragments were recovered by demineralisation, enzymatic digestion and LC-MS/MS on the Shenzhen Research Center's high-resolution mass spectrometry suite, identified by database search against the library alignment combined with de novo sequencing, and controlled at 1% FDR at the peptide-spectrum match level (Cox and Mann, 2008). Every batch included extraction blanks and procedural blanks. Candidate fragments were screened against the contamination exclusion database (human keratins, porcine trypsin, bovine collagen and laboratory-adjacent avian collagens) and were required to show deamidation consistent with specimen age. Clade assignment follows the existing procedure, in which specimen attribution and fragment placement are combined, and fragments are assigned at the level of Theropoda (JM-TR-2019-004).

Table 2. Sources of theropod-assigned fragments added in Library v3.2.
Source Description Fragments added Of which <8 residues Unique residues added
JM-SP-0074 Isolated tooth, Kem Kem Group, acquired 2025 168 81 552
JM-SP-0075 Isolated tooth, Kem Kem Group, acquired 2025 121 58 401
Re-extraction Archived specimens including the founding specimen JM-SP-0001, revised demineralisation 185 72 497
Total 474 211 1,450

Both new specimens were acquired through the same commercial channel as earlier Kem Kem material and passed intake under the chain-of-custody schema (JM-TR-2021-005). Twelve candidate fragments from JM-SP-0075 were excluded as closer matches to bovine collagen. A further four were excluded because they were detected in the procedural blank of the same batch.

3.3 Placement and replicate variance

Every theropod-assigned fragment was placed on the v3.1 and v3.2 reference trees by maximum-likelihood placement with posterior probabilities computed from likelihood weights. To estimate placement variance, each fragment was placed twenty times per library version, each replicate using a different random seed and a nonparametric bootstrap resample of the reference alignment. For each fragment we recorded the standard deviation of its placement posterior across replicates and its flip rate: the proportion of replicates in which the clade receiving the greatest posterior mass differed from the clade so identified in the consensus run. Clades were defined at the level used for assignment: non-avian Theropoda, crown Neognathae, crown Palaeognathae, Crocodylia and its fossil relatives, or unresolved (no clade above 0.5).

3.4 Threshold removal analysis

To assess whether short fragments should be removed, we simulated minimum-length thresholds of 6, 7, 8, 9 and 10 residues. For each we recorded the fragments removed, the unique residue positions left without any fossil anchor, the effect on the median placement posterior, and the number of positions in the Deinonychus target whose inferred residue changes when the target is re-derived without the removed fragments.

4. Results

4.1 Composition of the v3.2 fragment set

The v3.2 theropod set contains 3,377 fragments with a median length of 11 residues. 735 fragments (21.8%) are shorter than 8 residues, up from 524 of 2,903 (18.1%) in v3.1 (Fig. 2). Short fragments contribute disproportionately few unique residues because they overlap longer fragments from the same region; 2,104 of the 17,410 unique residues (12.1%) are covered only by short fragments.

Figure 2. Length distribution of theropod-assigned fragments, Library v3.2Fragment length in residues for all 3,377 theropod-assigned fossil-derived fragments in v3.2. 735 fragments (21.8%) are shorter than 8 residues.Fragments010020030040050060070080056781012162030505–6: 966–7: 2387–8: 4018–10: 61210–12: 54812–16: 70116–20: 41820–30: 30130–50: 62Fragment length (residues)8-residue thresholdFigure 2. Length distribution of theropod-assigned fragments, Library v3.2Fragment length in residues for all 3,377 theropod-assigned fossil-derived fragments in v3.2. 735 fragments (21.8%) are shorter than 8 residues.Fragments020040060080010203040505–6: 966–7: 2387–8: 4018–10: 61210–12: 54812–16: 70116–20: 41820–30: 30130–50: 62Fragment length (residues)8-residuethreshold
Figure 2.Length distribution of theropod-assigned fragments, Library v3.2. Fragment length in residues for all 3,377 theropod-assigned fossil-derived fragments in v3.2. 735 fragments (21.8%) are shorter than 8 residues.Source: Library v3.2 freeze, entry metadata.
Source data
Fragment length (residues)Fragments
5–696
6–7238
7–8401
8–10612
10–12548
12–16701
16–20418
20–30301
30–5062

4.2 Placement: overall and by length

The median theropod placement posterior rose from 0.74 to 0.76. Among a stratified sample of 48 fragments placed under both versions (Fig. 3), 25 of 26 fragments of 8 residues or more increased in posterior, with one unchanged, and 21 of 22 shorter fragments decreased. The two groups move in opposite directions, and the overall median rises because longer fragments are the majority.

Figure 3. Placement posterior under v3.1 and v3.2Theropod placement posterior of 48 fragments sampled across the length distribution, placed under Library v3.1 (x) and v3.2 (y). Points above the diagonal improved. Linear fit across all points: slope 1.16, R² = 0.95.Placement posterior, v3.20.00.20.40.60.81.00.00.10.20.30.40.50.60.70.80.91.0Placement posterior, v3.1<8 residues linear fit: y = 0.873x − 0.0209, R² = 0.91≥8 residues linear fit: y = 0.984x + 0.0476, R² = 0.98<8 residues Placement posterior, v3.1: 0.21 Placement posterior, v3.2: 0.15<8 residues Placement posterior, v3.1: 0.58 Placement posterior, v3.2: 0.55<8 residues Placement posterior, v3.1: 0.4 Placement posterior, v3.2: 0.33<8 residues Placement posterior, v3.1: 0.4 Placement posterior, v3.2: 0.31<8 residues Placement posterior, v3.1: 0.55 Placement posterior, v3.2: 0.36<8 residues Placement posterior, v3.1: 0.57 Placement posterior, v3.2: 0.48<8 residues Placement posterior, v3.1: 0.34 Placement posterior, v3.2: 0.35<8 residues Placement posterior, v3.1: 0.24 Placement posterior, v3.2: 0.17<8 residues Placement posterior, v3.1: 0.18 Placement posterior, v3.2: 0.1<8 residues Placement posterior, v3.1: 0.55 Placement posterior, v3.2: 0.38<8 residues Placement posterior, v3.1: 0.6 Placement posterior, v3.2: 0.54<8 residues Placement posterior, v3.1: 0.23 Placement posterior, v3.2: 0.18<8 residues Placement posterior, v3.1: 0.61 Placement posterior, v3.2: 0.56<8 residues Placement posterior, v3.1: 0.19 Placement posterior, v3.2: 0.14<8 residues Placement posterior, v3.1: 0.2 Placement posterior, v3.2: 0.18<8 residues Placement posterior, v3.1: 0.25 Placement posterior, v3.2: 0.19<8 residues Placement posterior, v3.1: 0.32 Placement posterior, v3.2: 0.28<8 residues Placement posterior, v3.1: 0.54 Placement posterior, v3.2: 0.46<8 residues Placement posterior, v3.1: 0.5 Placement posterior, v3.2: 0.37<8 residues Placement posterior, v3.1: 0.52 Placement posterior, v3.2: 0.42<8 residues Placement posterior, v3.1: 0.44 Placement posterior, v3.2: 0.38<8 residues Placement posterior, v3.1: 0.29 Placement posterior, v3.2: 0.26≥8 residues Placement posterior, v3.1: 0.71 Placement posterior, v3.2: 0.76≥8 residues Placement posterior, v3.1: 0.58 Placement posterior, v3.2: 0.61≥8 residues Placement posterior, v3.1: 0.74 Placement posterior, v3.2: 0.78≥8 residues Placement posterior, v3.1: 0.83 Placement posterior, v3.2: 0.84≥8 residues Placement posterior, v3.1: 0.59 Placement posterior, v3.2: 0.65≥8 residues Placement posterior, v3.1: 0.57 Placement posterior, v3.2: 0.61≥8 residues Placement posterior, v3.1: 0.67 Placement posterior, v3.2: 0.69≥8 residues Placement posterior, v3.1: 0.56 Placement posterior, v3.2: 0.6≥8 residues Placement posterior, v3.1: 0.55 Placement posterior, v3.2: 0.61≥8 residues Placement posterior, v3.1: 0.93 Placement posterior, v3.2: 0.95≥8 residues Placement posterior, v3.1: 0.75 Placement posterior, v3.2: 0.82≥8 residues Placement posterior, v3.1: 0.8 Placement posterior, v3.2: 0.82≥8 residues Placement posterior, v3.1: 0.76 Placement posterior, v3.2: 0.78≥8 residues Placement posterior, v3.1: 0.7 Placement posterior, v3.2: 0.73≥8 residues Placement posterior, v3.1: 0.84 Placement posterior, v3.2: 0.88≥8 residues Placement posterior, v3.1: 0.75 Placement posterior, v3.2: 0.82≥8 residues Placement posterior, v3.1: 0.79 Placement posterior, v3.2: 0.83≥8 residues Placement posterior, v3.1: 0.61 Placement posterior, v3.2: 0.61≥8 residues Placement posterior, v3.1: 0.65 Placement posterior, v3.2: 0.68≥8 residues Placement posterior, v3.1: 0.93 Placement posterior, v3.2: 0.96≥8 residues Placement posterior, v3.1: 0.79 Placement posterior, v3.2: 0.85≥8 residues Placement posterior, v3.1: 0.68 Placement posterior, v3.2: 0.7≥8 residues Placement posterior, v3.1: 0.94 Placement posterior, v3.2: 0.98≥8 residues Placement posterior, v3.1: 0.58 Placement posterior, v3.2: 0.6≥8 residues Placement posterior, v3.1: 0.68 Placement posterior, v3.2: 0.72≥8 residues Placement posterior, v3.1: 0.58 Placement posterior, v3.2: 0.63Figure 3. Placement posterior under v3.1 and v3.2Theropod placement posterior of 48 fragments sampled across the length distribution, placed under Library v3.1 (x) and v3.2 (y). Points above the diagonal improved. Linear fit across all points: slope 1.16, R² = 0.95.Placement posterior, v3.20.00.20.40.60.81.00.00.20.40.60.81.0Placement posterior, v3.1<8 residues linear fit: y = 0.873x − 0.0209, R² = 0.91≥8 residues linear fit: y = 0.984x + 0.0476, R² = 0.98<8 residues Placement posterior, v3.1: 0.21 Placement posterior, v3.2: 0.15<8 residues Placement posterior, v3.1: 0.58 Placement posterior, v3.2: 0.55<8 residues Placement posterior, v3.1: 0.4 Placement posterior, v3.2: 0.33<8 residues Placement posterior, v3.1: 0.4 Placement posterior, v3.2: 0.31<8 residues Placement posterior, v3.1: 0.55 Placement posterior, v3.2: 0.36<8 residues Placement posterior, v3.1: 0.57 Placement posterior, v3.2: 0.48<8 residues Placement posterior, v3.1: 0.34 Placement posterior, v3.2: 0.35<8 residues Placement posterior, v3.1: 0.24 Placement posterior, v3.2: 0.17<8 residues Placement posterior, v3.1: 0.18 Placement posterior, v3.2: 0.1<8 residues Placement posterior, v3.1: 0.55 Placement posterior, v3.2: 0.38<8 residues Placement posterior, v3.1: 0.6 Placement posterior, v3.2: 0.54<8 residues Placement posterior, v3.1: 0.23 Placement posterior, v3.2: 0.18<8 residues Placement posterior, v3.1: 0.61 Placement posterior, v3.2: 0.56<8 residues Placement posterior, v3.1: 0.19 Placement posterior, v3.2: 0.14<8 residues Placement posterior, v3.1: 0.2 Placement posterior, v3.2: 0.18<8 residues Placement posterior, v3.1: 0.25 Placement posterior, v3.2: 0.19<8 residues Placement posterior, v3.1: 0.32 Placement posterior, v3.2: 0.28<8 residues Placement posterior, v3.1: 0.54 Placement posterior, v3.2: 0.46<8 residues Placement posterior, v3.1: 0.5 Placement posterior, v3.2: 0.37<8 residues Placement posterior, v3.1: 0.52 Placement posterior, v3.2: 0.42<8 residues Placement posterior, v3.1: 0.44 Placement posterior, v3.2: 0.38<8 residues Placement posterior, v3.1: 0.29 Placement posterior, v3.2: 0.26≥8 residues Placement posterior, v3.1: 0.71 Placement posterior, v3.2: 0.76≥8 residues Placement posterior, v3.1: 0.58 Placement posterior, v3.2: 0.61≥8 residues Placement posterior, v3.1: 0.74 Placement posterior, v3.2: 0.78≥8 residues Placement posterior, v3.1: 0.83 Placement posterior, v3.2: 0.84≥8 residues Placement posterior, v3.1: 0.59 Placement posterior, v3.2: 0.65≥8 residues Placement posterior, v3.1: 0.57 Placement posterior, v3.2: 0.61≥8 residues Placement posterior, v3.1: 0.67 Placement posterior, v3.2: 0.69≥8 residues Placement posterior, v3.1: 0.56 Placement posterior, v3.2: 0.6≥8 residues Placement posterior, v3.1: 0.55 Placement posterior, v3.2: 0.61≥8 residues Placement posterior, v3.1: 0.93 Placement posterior, v3.2: 0.95≥8 residues Placement posterior, v3.1: 0.75 Placement posterior, v3.2: 0.82≥8 residues Placement posterior, v3.1: 0.8 Placement posterior, v3.2: 0.82≥8 residues Placement posterior, v3.1: 0.76 Placement posterior, v3.2: 0.78≥8 residues Placement posterior, v3.1: 0.7 Placement posterior, v3.2: 0.73≥8 residues Placement posterior, v3.1: 0.84 Placement posterior, v3.2: 0.88≥8 residues Placement posterior, v3.1: 0.75 Placement posterior, v3.2: 0.82≥8 residues Placement posterior, v3.1: 0.79 Placement posterior, v3.2: 0.83≥8 residues Placement posterior, v3.1: 0.61 Placement posterior, v3.2: 0.61≥8 residues Placement posterior, v3.1: 0.65 Placement posterior, v3.2: 0.68≥8 residues Placement posterior, v3.1: 0.93 Placement posterior, v3.2: 0.96≥8 residues Placement posterior, v3.1: 0.79 Placement posterior, v3.2: 0.85≥8 residues Placement posterior, v3.1: 0.68 Placement posterior, v3.2: 0.7≥8 residues Placement posterior, v3.1: 0.94 Placement posterior, v3.2: 0.98≥8 residues Placement posterior, v3.1: 0.58 Placement posterior, v3.2: 0.6≥8 residues Placement posterior, v3.1: 0.68 Placement posterior, v3.2: 0.72≥8 residues Placement posterior, v3.1: 0.58 Placement posterior, v3.2: 0.63
Figure 3.Placement posterior under v3.1 and v3.2. Theropod placement posterior of 48 fragments sampled across the length distribution, placed under Library v3.1 (x) and v3.2 (y). Points above the diagonal improved. Linear fit across all points: slope 1.16, R² = 0.95.Source: Library v3.1 and v3.2, placement runs PL-2025-10 and PL-2026-07.
Source data
SeriesPlacement posterior, v3.1Placement posterior, v3.2
<8 residues0.210.15
<8 residues0.580.55
<8 residues0.40.33
<8 residues0.40.31
<8 residues0.550.36
<8 residues0.570.48
<8 residues0.340.35
<8 residues0.240.17
<8 residues0.180.1
<8 residues0.550.38
<8 residues0.60.54
<8 residues0.230.18
<8 residues0.610.56
<8 residues0.190.14
<8 residues0.20.18
<8 residues0.250.19
<8 residues0.320.28
<8 residues0.540.46
<8 residues0.50.37
<8 residues0.520.42
<8 residues0.440.38
<8 residues0.290.26
≥8 residues0.710.76
≥8 residues0.580.61
≥8 residues0.740.78
≥8 residues0.830.84
≥8 residues0.590.65
≥8 residues0.570.61
≥8 residues0.670.69
≥8 residues0.560.6
≥8 residues0.550.61
≥8 residues0.930.95
≥8 residues0.750.82
≥8 residues0.80.82
≥8 residues0.760.78
≥8 residues0.70.73
≥8 residues0.840.88
≥8 residues0.750.82
≥8 residues0.790.83
≥8 residues0.610.61
≥8 residues0.650.68
≥8 residues0.930.96
≥8 residues0.790.85
≥8 residues0.680.7
≥8 residues0.940.98
≥8 residues0.580.6
≥8 residues0.680.72
≥8 residues0.580.63

One short fragment in Fig. 3 ([0.34, 0.35]) increased marginally. It is the only short fragment in the sample that does not lie in a Gly-X-Y collagen repeat.

4.3 Replicate variance

Replicate standard deviation of the placement posterior falls steeply with length in both versions (Fig. 4). Above 8 residues the two versions are indistinguishable. Below 8 residues v3.2 is more variable at every length: at 7 residues the median standard deviation is 0.17 against 0.13 in v3.1, and at 5 residues 0.24 against 0.19.

Figure 4. Replicate variance of placement posterior by fragment lengthMedian standard deviation of the placement posterior across 20 replicate placements per fragment, by fragment length, for all theropod-assigned fragments. Band: interquartile range, v3.2. v3.1 n = 2,903; v3.2 n = 3,377.Replicate SD of placement posterior0.000.050.100.150.200.250.300.350.40681012141618202224Fragment length (residues)8-residue thresholdFragment length (residues): 5 v3.1: 0.19 v3.2: 0.24 (0.15–0.33)Fragment length (residues): 6 v3.1: 0.16 v3.2: 0.21 (0.13–0.29)Fragment length (residues): 7 v3.1: 0.13 v3.2: 0.17 (0.10–0.24)Fragment length (residues): 8 v3.1: 0.09 v3.2: 0.10 (0.06–0.14)Fragment length (residues): 9 v3.1: 0.07 v3.2: 0.07 (0.04–0.11)Fragment length (residues): 10 v3.1: 0.06 v3.2: 0.06 (0.04–0.09)Fragment length (residues): 12 v3.1: 0.05 v3.2: 0.05 (0.03–0.07)Fragment length (residues): 14 v3.1: 0.04 v3.2: 0.04 (0.03–0.06)Fragment length (residues): 16 v3.1: 0.04 v3.2: 0.04 (0.02–0.05)Fragment length (residues): 20 v3.1: 0.03 v3.2: 0.03 (0.02–0.04)Fragment length (residues): 24 v3.1: 0.03 v3.2: 0.03 (0.01–0.04)Figure 4. Replicate variance of placement posterior by fragment lengthMedian standard deviation of the placement posterior across 20 replicate placements per fragment, by fragment length, for all theropod-assigned fragments. Band: interquartile range, v3.2. v3.1 n = 2,903; v3.2 n = 3,377.Replicate SD of placement posterior0.000.100.200.300.405101520Fragment length (residues)8-residuethresholdFragment length (residues): 5 v3.1: 0.19 v3.2: 0.24 (0.15–0.33)Fragment length (residues): 6 v3.1: 0.16 v3.2: 0.21 (0.13–0.29)Fragment length (residues): 7 v3.1: 0.13 v3.2: 0.17 (0.10–0.24)Fragment length (residues): 8 v3.1: 0.09 v3.2: 0.10 (0.06–0.14)Fragment length (residues): 9 v3.1: 0.07 v3.2: 0.07 (0.04–0.11)Fragment length (residues): 10 v3.1: 0.06 v3.2: 0.06 (0.04–0.09)Fragment length (residues): 12 v3.1: 0.05 v3.2: 0.05 (0.03–0.07)Fragment length (residues): 14 v3.1: 0.04 v3.2: 0.04 (0.03–0.06)Fragment length (residues): 16 v3.1: 0.04 v3.2: 0.04 (0.02–0.05)Fragment length (residues): 20 v3.1: 0.03 v3.2: 0.03 (0.02–0.04)Fragment length (residues): 24 v3.1: 0.03 v3.2: 0.03 (0.01–0.04)
Figure 4.Replicate variance of placement posterior by fragment length. Median standard deviation of the placement posterior across 20 replicate placements per fragment, by fragment length, for all theropod-assigned fragments. Band: interquartile range, v3.2. v3.1 n = 2,903; v3.2 n = 3,377.Source: Library v3.1 and v3.2, replicate placement set PL-2026-07R.
Source data
Fragment length (residues)v3.1v3.2v3.2 interval
50.190.240.15–0.33
60.160.210.13–0.29
70.130.170.1–0.24
80.090.10.06–0.14
90.070.0750.045–0.11
100.060.0620.037–0.09
120.050.050.03–0.07
140.040.040.025–0.06
160.0350.0350.02–0.05
200.030.030.018–0.043
240.0250.0250.015–0.036

Table 3. Placement variance by length class, Library v3.1 and v3.2.
Length class v3.1 fragments v3.2 fragments Median replicate SD, v3.1 Median replicate SD, v3.2 Flip rate, v3.1 Flip rate, v3.2
5–7 residues 524 735 0.16 0.20 0.22 0.31
8–11 residues 1,014 1,160 0.07 0.07 0.05 0.05
12–19 residues 1,013 1,119 0.04 0.04 0.02 0.02
≥20 residues 352 363 0.03 0.03 0.01 0.01
All 2,903 3,377 0.06 0.06 0.06 0.09

A flip rate of 0.31 means that, for a typical short fragment, roughly one replicate in three places it in a different clade from the consensus.

4.4 Where short fragments go

The distribution of consensus placements by length shows where the variance comes from (Fig. 5). Among fragments of 16 residues or more, 90% place in non-avian Theropoda. Among 5-residue fragments, 28% do; 35% are unresolved, and the remaining mass is spread across crown birds and crocodilians. Fourteen per cent of 5-residue fragments now place in crown Palaeognathae, against 4% under v3.1.

Figure 5. Consensus placement clade by fragment length, Library v3.2Proportion of theropod-assigned fragments in each length class whose consensus placement (greatest posterior mass) falls in each clade under v3.2. Rows sum to 1. Unresolved: no clade above 0.5.Consensus placement cladeNon-avian TheropodaNeognathaePalaeognathaeCrocodylia…Unresolved55 × Non-avian Theropoda: 0.280.285 × Neognathae: 0.170.175 × Palaeognathae: 0.140.145 × Crocodylia and relatives: 0.060.065 × Unresolved: 0.350.3566 × Non-avian Theropoda: 0.360.366 × Neognathae: 0.150.156 × Palaeognathae: 0.120.126 × Crocodylia and relatives: 0.050.056 × Unresolved: 0.320.3277 × Non-avian Theropoda: 0.450.457 × Neognathae: 0.130.137 × Palaeognathae: 0.10.107 × Crocodylia and relatives: 0.040.047 × Unresolved: 0.280.288–98–9 × Non-avian Theropoda: 0.680.688–9 × Neognathae: 0.080.088–9 × Palaeognathae: 0.040.048–9 × Crocodylia and relatives: 0.030.038–9 × Unresolved: 0.170.1710–1110–11 × Non-avian Theropoda: 0.770.7710–11 × Neognathae: 0.060.0610–11 × Palaeognathae: 0.020.0210–11 × Crocodylia and relatives: 0.020.0210–11 × Unresolved: 0.130.1312–1512–15 × Non-avian Theropoda: 0.840.8412–15 × Neognathae: 0.040.0412–15 × Palaeognathae: 0.010.0112–15 × Crocodylia and relatives: 0.010.0112–15 × Unresolved: 0.10.10≥16≥16 × Non-avian Theropoda: 0.90.90≥16 × Neognathae: 0.030.03≥16 × Palaeognathae: 00.00≥16 × Crocodylia and relatives: 0.010.01≥16 × Unresolved: 0.060.06Figure 5. Consensus placement clade by fragment length, Library v3.2Proportion of theropod-assigned fragments in each length class whose consensus placement (greatest posterior mass) falls in each clade under v3.2. Rows sum to 1. Unresolved: no clade above 0.5.Consensus placement cladeNon-avian TheropodaNeognathaePalaeognathaeCrocodylia…Unresolved55 × Non-avian Theropoda: 0.280.285 × Neognathae: 0.170.175 × Palaeognathae: 0.140.145 × Crocodylia and relatives: 0.060.065 × Unresolved: 0.350.3566 × Non-avian Theropoda: 0.360.366 × Neognathae: 0.150.156 × Palaeognathae: 0.120.126 × Crocodylia and relatives: 0.050.056 × Unresolved: 0.320.3277 × Non-avian Theropoda: 0.450.457 × Neognathae: 0.130.137 × Palaeognathae: 0.10.107 × Crocodylia and relatives: 0.040.047 × Unresolved: 0.280.288–98–9 × Non-avian Theropoda: 0.680.688–9 × Neognathae: 0.080.088–9 × Palaeognathae: 0.040.048–9 × Crocodylia and relatives: 0.030.038–9 × Unresolved: 0.170.1710–1110–11 × Non-avian Theropoda: 0.770.7710–11 × Neognathae: 0.060.0610–11 × Palaeognathae: 0.020.0210–11 × Crocodylia and relatives: 0.020.0210–11 × Unresolved: 0.130.1312–1512–15 × Non-avian Theropoda: 0.840.8412–15 × Neognathae: 0.040.0412–15 × Palaeognathae: 0.010.0112–15 × Crocodylia and relatives: 0.010.0112–15 × Unresolved: 0.10.10≥16≥16 × Non-avian Theropoda: 0.90.90≥16 × Neognathae: 0.030.03≥16 × Palaeognathae: 00.00≥16 × Crocodylia and relatives: 0.010.01≥16 × Unresolved: 0.060.06
Figure 5.Consensus placement clade by fragment length, Library v3.2. Proportion of theropod-assigned fragments in each length class whose consensus placement (greatest posterior mass) falls in each clade under v3.2. Rows sum to 1. Unresolved: no clade above 0.5.Source: Library v3.2, placement run PL-2026-07.
Source data
Fragment length (residues)Non-avian TheropodaNeognathaePalaeognathaeCrocodylia and relativesUnresolved
50.280.170.140.060.35
60.360.150.120.050.32
70.450.130.10.040.28
8–90.680.080.040.030.17
10–110.770.060.020.020.13
12–150.840.040.010.010.1
≥160.90.0300.010.06

In total, 82 short fragments have a consensus placement in crown Palaeognathae under v3.2, against 23 under v3.1. Of the 82, 57 are identical over their full length to the corresponding Struthio camelus reference sequence, and each of the 57 is also identical to at least 30 other avian references. At five to seven residues, most of them within the Gly-X-Y repeat of type I collagen, identity to an ostrich is identity to most birds, and the placement algorithm distributes posterior mass across all of them. Adding seventeen closely related palaeognaths adds seventeen more places for that mass to go. This dilution, rather than any change in the fragments themselves, accounts for most of the v3.1 to v3.2 increase in variance.

Because modern avian collagen is a recognised contamination risk for this kind of work (Buckley et al., 2008; Saitta et al., 2019), the palaeognath-placed fragments were reviewed by the palaeoproteomics group. All 82 show deamidation consistent with their specimens, none was detected in any extraction or procedural blank, and all come from batches in which the blanks were clean. No palaeognath material has been handled at the Shenzhen Research Center. The exclusion database contains Gallus collagen as a laboratory-adjacent avian reference; it does not contain palaeognath collagen, and we do not propose to add it, because doing so would exclude every fragment identical to it and therefore most short fragments in the library.

4.5 Effect on reconstructed targets

The Deinonychus target re-derived under v3.2 differs from its v3.1 counterpart at 7 of 11,842 positions; the median per-position posterior is unchanged at 0.93. The Brachiosaurus target differs at 12 of 13,273 positions; its median per-position posterior rises from 0.81 to 0.82, reflecting a better-constrained crown-bird node. Placing rheas as sister to all other non-ostrich palaeognaths, or tinamous in that position instead, changed 2 and 3 positions in the Deinonychus target respectively.

The production line DN-2019-04 was compared with the v3.2 Deinonychus target under the re-attestation procedure (JM-TR-2025-007). The expressed sequence now differs from the current reference target at 21 positions, the 14 recorded at v3.0 and seven more. All 21 are below a posterior of 0.6. The line remains within its acceptance criteria and continues to be attested to v3.0. Re-attestation to v3.2 is not required, and validation criteria are unchanged.

4.6 Threshold removal

Removing fragments below a minimum length reduces variance and raises the headline posterior, at the cost of coverage (Table 4, Fig. 6). At the 8-residue threshold, 735 fragments would be removed, the median placement posterior would rise from 0.76 to 0.81, and 412 unique residue positions would be left with no fossil anchor of any length. Re-deriving the Deinonychus target without them changes 31 of its 11,842 positions.

Table 4. Effect of minimum-length thresholds on the v3.2 theropod fragment set.
Minimum length Fragments removed Fragments remaining Positions left without fossil anchor Deinonychus target positions changed Median placement posterior Founding-specimen fragments removed
None (v3.2 as released) 0 3,377 0 0 0.76 0%
6 residues 96 3,281 38 2 0.77 6%
7 residues 334 3,043 151 9 0.79 19%
8 residues 735 2,642 412 31 0.81 41%
9 residues 1,048 2,329 688 52 0.82 52%
10 residues 1,347 2,030 1,027 87 0.83 60%

Figure 6. Fragments removed and anchors lost under minimum-length thresholdsFor each simulated minimum fragment length: theropod-assigned fragments removed, and unique residue positions left with no fossil anchor of any length. Library v3.2, n = 3,377 fragments, 17,410 unique residues.Count02004006008001,0001,2001,40066 Fragments removed: 966 Positions left without fossil anchor: 3877 Fragments removed: 3347 Positions left without fossil anchor: 15188 Fragments removed: 7358 Positions left without fossil anchor: 41299 Fragments removed: 1,0489 Positions left without fossil anchor: 6881010 Fragments removed: 1,34710 Positions left without fossil anchor: 1,027Minimum fragment length (residues)Figure 6. Fragments removed and anchors lost under minimum-length thresholdsFor each simulated minimum fragment length: theropod-assigned fragments removed, and unique residue positions left with no fossil anchor of any length. Library v3.2, n = 3,377 fragments, 17,410 unique residues.Count02004006008001,0001,2001,40066 Fragments removed: 966 Positions left without fossil anchor: 3877 Fragments removed: 3347 Positions left without fossil anchor: 15188 Fragments removed: 7358 Positions left without fossil anchor: 41299 Fragments removed: 1,0489 Positions left without fossil anchor: 6881010 Fragments removed: 1,34710 Positions left without fossil anchor: 1,027Minimum fragment length (residues)
Figure 6.Fragments removed and anchors lost under minimum-length thresholds. For each simulated minimum fragment length: theropod-assigned fragments removed, and unique residue positions left with no fossil anchor of any length. Library v3.2, n = 3,377 fragments, 17,410 unique residues.Source: Library v3.2, threshold simulation PL-2026-07T.
Source data
Minimum fragment length (residues)Fragments removedPositions left without fossil anchor
69638
7334151
8735412
91,048688
101,3471,027

The founding specimen (JM-SP-0001) is disproportionately affected. Its fragments are on average shorter than those of later specimens, a consequence of its age at extraction, of the methods available in 2011–2012 (JM-TR-2013-001), and of repeated re-extraction. Removal at 8 residues would remove 41% of the fragments it has contributed to the library.

5. Discussion

The palaeognath expansion did what it was intended to do. The crown-bird node is better constrained, the Brachiosaurus target's median posterior improved for the first time since its profile was introduced, and fragments long enough to carry phylogenetic signal place more confidently. The release also made a known weakness worse. These are the same result viewed from opposite ends of the length distribution: a denser reference set rewards informative fragments and penalises uninformative ones, because it gives the former more precise targets and the latter more equally good places to land.

The case for removing short fragments is straightforward and was made internally. They are the main source of placement variance, they lower the headline posterior, and they have been on the known-issue register for seven years. Removal at 8 residues would raise the median placement posterior to 0.81, a figure the library would otherwise not be expected to reach for several releases.

We have not adopted it, for three reasons. First, variance is not bias. Replicate placements of short fragments are dispersed, but their consensus placements are not systematically wrong: across the full set, short- fragment consensus placements agree with those of overlapping longer fragments from the same specimen in 87% of cases where such overlap exists. Second, the cost of removal falls on coverage, and coverage is what the library is for. At 8 residues, 412 library positions would revert from fossil-anchored to reference-only, and 31 positions in the Deinonychus target would change. In reconstruction terms, a noisy direct observation constrains these positions more than none at all, provided its uncertainty is propagated, which since v2.8 it is (JM-TR-2023-006). Third, placement variance is already carried forward: placement posteriors weight each fragment's contribution to substitution-model calibration, so a high-variance fragment contributes correspondingly little. The variance is real, measured and accounted for.

We note, without drawing a conclusion from it, that the fragments that would be removed include 41% of those recovered from the specimen on which the program was founded.

The question will be revisited if a future release adds reference taxa that further dilute short-fragment placement, or if longer fragments come to cover the positions currently anchored only by short ones. Neither is expected in the next release. The known issue was presented to the Ethics & Bioresponsibility Council on 22 July 2026: reviewed; no concerns raised.

6. Limitations and known issues

  • Short-fragment variance (KI-0601; internal KI-2019-02). Fragments shorter than 8 residues remain high-variance in placement. In v3.2 the flip rate for this class is 0.31. Documented. Not scheduled for removal.
  • Palaeognath placements. 82 short theropod-assigned fragments have a consensus placement in crown Palaeognathae, and 57 are identical over their full length to the ostrich reference. They are retained as theropod-assigned on specimen attribution and authentication evidence. Their placements are uninformative at this length and are recorded as such.
  • Clade assignment for short fragments is specimen-driven. For fragments whose placement is unresolved, assignment to Theropoda rests on the attribution of the specimen. Specimen attribution for Kem Kem teeth is established at acquisition and refined at intake; it is not independent of the commercial channel.
  • Reference sequences are not palaeognath-curated. Added palaeognath sequences were assembled from public resources and reviewed for alignment, not re-sequenced. Coverage of the tinamous is lowest (93.2%).
  • Palaeognath topology is unsettled. Two alternative arrangements were tested (Section 4.5). Others were not.
  • Variance figures are replicate-based. The 20-replicate design estimates variance due to model and resampling. It does not capture variance due to the original identification of the fragment, which for short de novo sequences may be larger.
  • The headline median is length-weighted. The reported median placement posterior (0.76) includes short fragments. It would be 0.81 without them. Both figures are given here so that neither is reported alone.

7. Changes in Library v3.2

As published in the Peptide Library Release Notes:

  • Minor: updated avian reference set to include additional paleognath sequences.
  • Known issue: very short peptide fragments (<8 residues) remain high-variance in placement. Documented. Not scheduled for removal.
  • No changes to downstream cell-line validation criteria.

Also listed in the release notes for v3.2:

  • 474 theropod-assigned fossil-derived entries added (2,903 → 3,377); residue coverage 15,960 → 17,410.
  • Per-entry fragment-length field in the build manifest, and a reference-completeness flag for extant reference sequences below full length.
  • Short fragments (<8 residues) now carry reduced weight in bracketing, according to the per-entry placement variance (Section 5). The production line DN-2019-04 remains within its acceptance criteria and continues to reference v3.0 (Section 4.5).
  • Validation report now lists placement posterior distributions by fragment-length class.
  • Known issue: two newly added palaeognath taxa have partial COL1A2 references (≥ 92% length) (KI-0604).

Library v3.2 contains 3,377 theropod-assigned fossil-derived fragments covering 17,410 unique residues, with a median theropod placement posterior of 0.76, and reference sequences from 88 extant taxa (17 added, all palaeognaths). No sauropodomorph-assigned peptides are included. The reference expansion is described in Sections 3.1 and 4.5, the known issue in Sections 4.2 to 4.6, and the validation position in Section 4.5.

8. Data availability

Library contents are confidential. Placement runs (PL-2025-10, PL-2026-07), the replicate set (PL-2026-07R) and the threshold simulation (PL-2026-07T) are retained at the Shenzhen Research Center and are available to authorized research staff. Library access is not available to third parties.

9. Acknowledgements

We thank Ruoxi Deng for sample preparation and instrument time on the re-extraction series, and Dr. Rahul Venkataraman for re-deriving the Brachiosaurus target under v3.2. We thank a dealer in the Rissani Sunday market for assistance with specimen provenance for JM-SP-0074 and JM-SP-0075.

This work was funded internally. All authors are employees of Jurassic Meat and may hold equity. The known issue and the decision to retain short fragments were reviewed by the Ethics & Bioresponsibility Council (22 July 2026): reviewed; no concerns raised.

10. References

  1. Asara JM, et al. (2007). Protein sequences from mastodon and Tyrannosaurus rex revealed by mass spectrometry. Science 316:280–285.
  2. Buckley M, et al. (2008). Comment on "Protein sequences from mastodon and Tyrannosaurus rex revealed by mass spectrometry." Science 319:33.
  3. Demarchi B, et al. (2016). Protein sequences bound to mineral surfaces persist into deep time. eLife 5:e17092.
  4. Cappellini E, et al. (2019). Early Pleistocene enamel proteome from Dmanisi resolves Stephanorhinus phylogeny. Nature 574:103–107.
  5. Welker F, et al. (2019). Enamel proteome shows that Gigantopithecus was an early diverging pongine. Nature 576:262–265.
  6. Saitta ET, et al. (2019). Cretaceous dinosaur bone contains recent organic material and provides an environment conducive to microbial communities. eLife 8:e46205.
  7. Witmer LM (1995). The extant phylogenetic bracket and the importance of reconstructing soft tissues in fossils. In: Thomason JJ (ed.), Functional Morphology in Vertebrate Paleontology. Cambridge University Press, 19–33.
  8. Cox J, Mann M (2008). MaxQuant enables high peptide identification rates, individualized p.p.b.-range mass accuracies and proteome-wide protein quantification. Nature Biotechnology 26:1367–1372.
  9. JM-TR-2013-001. Endogenous peptide fragments recovered from a Kem Kem Group theropod tooth. Jurassic Meat internal report.
  10. JM-TR-2019-004. Library v2.0: a versioned multiple-sequence-alignment architecture for fragmentary ancient peptides. Jurassic Meat internal report.
  11. JM-TR-2021-005. Library v2.3: keratin-associated peptides and the chain-of-custody metadata schema. Jurassic Meat internal report.
  12. JM-TR-2023-006. Library v2.8: diagenetic modification in placement scoring. Jurassic Meat internal report.
  13. JM-TR-2025-007. Library v3.0: commit-hash provenance for cell-line constructions, and a retrospective audit of the cross-clade build-filter defect. Jurassic Meat internal report.
  14. JM-TR-2025-008. Library v3.1: a sauropodomorph build profile for the Brachiosaurus program, constrained by extant reference taxa and bracketing alone. Jurassic Meat internal report.

Supplementary Table S1

Table S1. Theropod-assigned fragment set by library version, v3.0–v3.2.
Library Theropod fragments <8 residues % <8 residues Unique residues Median placement posterior Extant reference taxa Palaeognath taxa
v3.0 2,903 524 18.1 15,960 0.74 63 5
v3.1 2,903 524 18.1 15,960 0.74 71 5
v3.2 3,377 735 21.8 17,410 0.76 88 22