申请配额 联系我们

科学技术报告

JM-TR-2021-005 技术报告 文库 v2.3

Library v2.3: low-confidence placement of keratin-associated peptides from ungual sheath material, and a chain-of-custody metadata schema for fossil-derived entries

Dr. Hao Lin 1,* Dr. Rahul Venkataraman 1 Marcus Oyelaran 1

1 侏罗纪肉业 深圳研究中心
* 通讯作者: Dr. Hao Lin · 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 v2.3
  9. 8. Data availability
  10. 9. Acknowledgements
  11. 10. References

摘要

Library v2.3 adds 326 theropod-assigned fragments from five newly accessioned specimens, bringing the library to 1,610 fragments, 9,410 unique residues covered and 41 extant reference taxa, with a median placement posterior of 0.66. For the first time the release includes keratin-associated peptides (KAPs), recovered from sheath material adhering to two ungual (claw) specimens. The 87 KAP fragments place poorly (median posterior 0.29). They fall into two classes: fragments matching archosaur corneous beta-proteins, whose placement is limited by paralogy within a multi-copy gene family, and fragments matching alpha-keratins, whose placement degrades with increasing identity to human keratins, a ubiquitous laboratory contaminant. Adding reference taxa does not improve placement of either class. KAPs are retained in default builds with a structured flag; they are not removed. In parallel, v2.3 introduces a mandatory chain-of-custody metadata schema for all fossil-derived entries, comprising 25 fields covering acquisition, locality, documentation, custody events, preparation history and analytical linkage. Records were backfilled for the nine specimens accessioned before the schema took effect. Six of 14 specimens, contributing 825 fragments (51.2%), carry documented custody gaps. Custody status is recorded and reported for every build; it does not determine inclusion.

关键词palaeoproteomicskeratincorneous beta-proteinscontamination controlphylogenetic placementchain of custodymetadata standardsspecimen provenance

JM-TR-2021-005 · Library v2.3 · 9 February 2021 Dr. Hao Lin*, Dr. Rahul Venkataraman, Marcus Oyelaran Shenzhen Research Center, Jurassic Meat, Nanshan District, Shenzhen *Corresponding author

Public summary of internal report (full internal version: 38 pages).

1. Summary

  • Library v2.3 contains 1,610 theropod-assigned fragments from 14 specimens, covering 9,410 unique residues, with 41 extant reference taxa. Median placement posterior is 0.66.
  • 87 keratin-associated peptides from two ungual specimens place with low confidence (median posterior 0.29). Increasing the reference set does not improve this. The fragments are flagged, not removed.
  • Placement of alpha-keratin-matching fragments declines linearly with identity to human keratins. Nine such fragments differ from a human keratin peptide only at residues consistent with deamidation.
  • A 25-field chain-of-custody schema is now mandatory for every fossil-derived entry. Builds fail if any fossil-derived fragment cannot be linked to a complete specimen record.
  • Six of 14 specimens, including the founding specimen, carry documented custody gaps. These specimens contribute 825 of 1,610 fragments.

2. Introduction

Library v2.0 (JM-TR-2019-004) established a versioned, locus-partitioned alignment architecture in which every fragment is accounted for in every build. Since then the library has grown through two internal builds (v2.1, v2.2) to the present release. The additions in v2.3 are notable for two reasons.

First, two of the five new specimens are unguals (claw cores) with adherent sheath material, from which we recovered peptides attributable to keratin-family proteins. Keratin-associated peptides are of interest because the corneous beta-proteins (CBPs, historically "beta-keratins") are restricted to reptiles and birds and have no mammalian homologues. A CBP peptide cannot therefore be explained by contamination from human skin, which is the most common source of keratin in any proteomics laboratory. The alpha-keratins, by contrast, are shared across amniotes, and alpha-keratin peptides from fossil material are by default suspect. The difficulty of authenticating deep-time protein, and of excluding modern contamination and microbial input, is well documented (Schweitzer et al. 2007; Asara et al. 2007; Buckley et al. 2008; Saitta et al. 2019), even where persistence in mineral-bound contexts is plausible (Demarchi et al. 2016), and we have approached these fragments with corresponding caution.

Second, the growth of the library from four specimens in v1.4 to fourteen in v2.3 has exposed the limitations of the specimen-level documentation inherited from the company's early years. The intake and custody practice described in JM-TR-2016-003 (SOP-SP-003, with provenance tiers A–C) was recorded in the specimen register, in paper files and in free-text accession notes. Library entries carried the specimen ID only, and linking an entry to its custody record required a manual lookup (JM-TR-2016-003, Section 6). The v2.0 post-mortem (JM-TR-2019-004, Section 6) noted that the record for the founding specimen was incomplete for the period before 2011. Library v2.3 introduces a structured schema to address this, and backfills it for all existing specimens.

3. Materials and Methods

3.1 New specimens

Five specimens were accessioned between November 2019 and October 2020 under SOP-SP-003 (JM-TR-2016-003): JM-SP-0041 (isolated tooth), JM-SP-0042 and JM-SP-0045 (long-bone fragments), all Tier A, and JM-SP-0043 and JM-SP-0044 (manual unguals with adherent sheath material), both Tier B. The unguals were received with sheath material partly consolidated. Consolidants were characterised before sampling (Section 3.5), and sampling targeted unconsolidated regions on the ventral surface. All specimens were sampled in the clean-room suite under the procedures of JM-TR-2016-003, with extraction blanks in every batch.

3.2 Extraction and LC-MS/MS

Bone and tooth specimens were processed as in JM-TR-2019-004. Sheath material was processed separately, without demineralisation, using a reducing and chaotropic extraction appropriate to keratinous tissue, followed by tryptic digestion. All sheath extractions were performed in dedicated batches with two procedural blanks per batch, on days when no other keratinous sample was handled in the suite. Analysts wore full clean-room garments, including hoods, for all sheath work.

3.3 Identification and contamination screening

Spectra were searched against the v2.3 reference set, augmented with CBP and alpha-keratin sequences from 14 extant reptiles and birds, and against the contamination database (human keratins and keratin-associated proteins, porcine trypsin, bovine and porcine collagens, common laboratory proteins). Variable modifications, MaxQuant-type processing (Cox & Mann 2008) and acceptance criteria (1% peptide-level FDR) were as in JM-TR-2019-004. The standing contamination rule was applied without change: a peptide matching the contamination database at one substitution or fewer, or present in any blank from the same batch, is excluded and logged. Deamidation-consistent differences (N/D, Q/E) are counted as substitutions under this rule, as in all previous releases.

3.4 Classification of keratin-associated peptides

Fragments whose best match in the reference set was a keratin-family protein were designated KAPs and assigned to two classes: class A, best match to a CBP; and class B, best match to an alpha-keratin. For each class B fragment we recorded percentage identity to its nearest human keratin or keratin-associated protein over the aligned length.

3.5 Chain-of-custody schema and backfill

The schema (Table 4) was drafted by the laboratory operations and palaeoproteomics groups between March and July 2020 and adopted on 1 July 2020. Records for the nine specimens accessioned before that date were backfilled from accession files, purchase records, correspondence, shipping documents and laboratory notebooks, and marked provenance_backfilled = true. Where a fact could not be established, the field was recorded as unknown rather than left empty, and a custody gap was opened. Backfill included a conversation with the vendor of the founding specimen (Section 9).

3.6 Placement and reference sensitivity

Placement followed JM-TR-2019-004. To test whether poor KAP placement reflects sparse reference sampling, we re-ran placement on reference subsets of 12, 19, 26, 34 and 41 taxa, drawn to preserve the proportion of avian, crocodilian and outgroup taxa, with 20 random subsets per size.

4. Results

4.1 Library content

Library v2.3 contains 1,610 theropod-assigned fragments, an increase of 326 over v2.0, all from the five new specimens (Table 1, Fig. 1). Unique residues covered rose from 7,920 to 9,410 and extant reference taxa from 34 to 41. The 87 KAPs constitute 5.4% of the library and come exclusively from the two ungual specimens, which also contributed six collagen fragments from the bone core.

Table 1. Library statistics, v2.0 and v2.3.
Metric v2.0 (2019-06) v2.3 (2021-02)
Fossil specimens 9 14
Theropod-assigned fragments 1,284 1,610
— collagen type I 1,021 1,219
— enamel proteins 182 210
— other non-collagenous proteins 81 94
— keratin-associated (flagged) 0 87
Unique residues covered 7,920 9,410
Extant reference taxa 34 41
Median placement posterior (theropod), all fragments 0.63 0.66
Median placement posterior, excluding flagged fragments — 0.67
Fragments excluded by logged rule 17 26

Figure 1. Library composition by protein class, v2.0 and v2.3Theropod-assigned fragments by protein class. Keratin-associated peptides first appear in v2.3, all from specimens JM-SP-0043 and JM-SP-0044.Fragments02004006008001,0001,2001,4001,6001,800v2.0v2.0 Collagen type I: 1,021 Total: 1,284v2.0 Enamel proteins: 182 Total: 1,284v2.0 Other NCP: 81 Total: 1,2841,284v2.3v2.3 Collagen type I: 1,219 Total: 1,610v2.3 Enamel proteins: 210 Total: 1,610v2.3 Other NCP: 94 Total: 1,610v2.3 Keratin-associated (flagged): 87 Total: 1,6101,610Library versionFigure 1. Library composition by protein class, v2.0 and v2.3Theropod-assigned fragments by protein class. Keratin-associated peptides first appear in v2.3, all from specimens JM-SP-0043 and JM-SP-0044.Fragments05001,0001,5002,000v2.0v2.0 Collagen type I: 1,021 Total: 1,284v2.0 Enamel proteins: 182 Total: 1,284v2.0 Other NCP: 81 Total: 1,2841,284v2.3v2.3 Collagen type I: 1,219 Total: 1,610v2.3 Enamel proteins: 210 Total: 1,610v2.3 Other NCP: 94 Total: 1,610v2.3 Keratin-associated (flagged): 87 Total: 1,6101,610Library version
Figure 1.Library composition by protein class, v2.0 and v2.3. Theropod-assigned fragments by protein class. Keratin-associated peptides first appear in v2.3, all from specimens JM-SP-0043 and JM-SP-0044.Source: JM Shenzhen Research Center, build manifest v2.3.
Source data
Library versionCollagen type IEnamel proteinsOther NCPKeratin-associated (flagged)
v2.01,021182810
v2.31,2192109487

4.2 Keratin-associated peptides

Of the 87 KAPs, 54 are class A (CBP) and 33 are class B (alpha-keratin). Their placement posteriors are low and broadly distributed (Fig. 2; Table 2). Excluding KAPs from the headline calculation would raise the library median from 0.66 to 0.67. We report the figure with KAPs included.

Table 2. Keratin-associated peptides in Library v2.3.
Class A (CBP) Class B (alpha-keratin) All KAPs
Fragments 54 33 87
Median length (residues) 11 13 12
Median placement posterior 0.34 0.20 0.29
Fragments with posterior ≥ 0.5 14 2 16
Fragments carrying annotated deamidation 21 17 38
Has mammalian homologue No Yes —
Nearest-human-keratin identity, range n/a 62–93% —
Flag applied LOWCONF_KAP LOWCONF_KAP —

Figure 2. Placement posterior of keratin-associated peptides, Library v2.3Posterior placement probability on the avian stem for all 87 KAP fragments (classes A and B combined). Median 0.29. For comparison, the median for non-KAP fragments is 0.67.Fragments051015200.00.10.20.30.40.50.60.70.80.91.00–0.1: 140.1–0.2: 190.2–0.3: 150.3–0.4: 130.4–0.5: 100.5–0.6: 70.6–0.7: 50.7–0.8: 30.8–0.9: 1Posterior placement probabilityFigure 2. Placement posterior of keratin-associated peptides, Library v2.3Posterior placement probability on the avian stem for all 87 KAP fragments (classes A and B combined). Median 0.29. For comparison, the median for non-KAP fragments is 0.67.Fragments051015200.00.20.40.60.81.00–0.1: 140.1–0.2: 190.2–0.3: 150.3–0.4: 130.4–0.5: 100.5–0.6: 70.6–0.7: 50.7–0.8: 30.8–0.9: 1Posterior placement probability
Figure 2.Placement posterior of keratin-associated peptides, Library v2.3. Posterior placement probability on the avian stem for all 87 KAP fragments (classes A and B combined). Median 0.29. For comparison, the median for non-KAP fragments is 0.67.Source: JM Shenzhen Research Center, build manifest v2.3.
Source data
Posterior placement probabilityFragments
0–0.114
0.1–0.219
0.2–0.315
0.3–0.413
0.4–0.510
0.5–0.67
0.6–0.75
0.7–0.83
0.8–0.91
0.9–10

Class A. CBPs are encoded by a cluster of tandemly duplicated genes whose copies have been subject to repeated duplication, loss and sequence homogenisation. A short CBP peptide frequently matches several paralogues in each reference taxon equally well, and the placement engine, which operates on a locus-level alignment, cannot resolve which paralogue is orthologous to the fossil source. Posterior mass is consequently distributed across the reference topology rather than concentrated on the avian stem (Fig. 3). In 9 of 54 class A fragments the maximum-posterior placement is within crocodilians.

Figure 3. Distribution of placement mass for class A (CBP) fragmentsMean posterior placement mass across the 54 class A fragments, summarised on a simplified reference topology (11 of 41 extant reference taxa shown). Mass not shown is distributed across internal branches. Compare with the concentrated placement of non-KAP fragments (JM-TR-2019-004, Fig. 3).JM theropodp = 0.34JM theropod — placement posterior 0.34StruthioStruthioDromaiusDromaiusGallusp = 0.09Gallus — placement posterior 0.09AnasAnasTaeniopygiap = 0.06Taeniopygia — placement posterior 0.06Alligatorp = 0.08Alligator — placement posterior 0.08CaimanCaimanCrocodylusp = 0.05Crocodylus — placement posterior 0.05GavialisGavialisChrysemysp = 0.04Chrysemys — placement posterior 0.04PelodiscusPelodiscusFigure 3. Distribution of placement mass for class A (CBP) fragmentsMean posterior placement mass across the 54 class A fragments, summarised on a simplified reference topology (11 of 41 extant reference taxa shown). Mass not shown is distributed across internal branches. Compare with the concentrated placement of non-KAP fragments (JM-TR-2019-004, Fig. 3).JM theropodp = 0.34JM theropod — placement posterior 0.34StruthioStruthioDromaiusDromaiusGallusp = 0.09Gallus — placement posterior 0.09AnasAnasTaeniopygiap = 0.06Taeniopygia — placement posterior 0.06Alligatorp = 0.08Alligator — placement posterior 0.08CaimanCaimanCrocodylusp = 0.05Crocodylus — placement posterior 0.05GavialisGavialisChrysemysp = 0.04Chrysemys — placement posterior 0.04PelodiscusPelodiscus
Figure 3.Distribution of placement mass for class A (CBP) fragments. Mean posterior placement mass across the 54 class A fragments, summarised on a simplified reference topology (11 of 41 extant reference taxa shown). Mass not shown is distributed across internal branches. Compare with the concentrated placement of non-KAP fragments (JM-TR-2019-004, Fig. 3).Source: JM Shenzhen Research Center, build manifest v2.3.
Source data
TaxonHighlightedPlacement posterior
JM theropodyes0.34
Struthio
Dromaius
Gallus0.09
Anas
Taeniopygia0.06
Alligator0.08
Caiman
Crocodylus0.05
Gavialis
Chrysemys0.04
Pelodiscus

(((JM_theropod,((Struthio,Dromaius),((Gallus,Anas),Taeniopygia))),((Alligator,Caiman),(Crocodylus,Gavialis))),(Chrysemys,Pelodiscus));

Class B. Placement posterior for class B fragments declines with identity to the nearest human keratin (Fig. 4; slope −0.0098 per percentage point, R² = 0.88). This relationship is expected if the fragments, or the spectra from which they were identified, contain a contribution from human keratin, and it is also expected if the fragments are endogenous and simply conserved: a highly conserved alpha-keratin peptide is both close to the human sequence and uninformative about placement. The two explanations cannot be distinguished on placement alone.

Nine class B fragments, at 82–91% identity, differ from a human keratin peptide at exactly two positions, both of which are consistent with deamidation (N→D or Q→E). Under the standing contamination rule, which counts deamidation-consistent differences as substitutions, these fragments match the contamination database at two substitutions and are retained. If the differences were treated as modifications, they would match at zero substitutions. We note that deamidation is the modification on which the authentication of our oldest fragments has relied since JM-TR-2013-001. Four further class B fragments matched a human keratin at one substitution, were excluded by rule, and were restored by curator override (SUPPRESS_CONTAM) after review of their spectra; the rationale is recorded in each override.

Figure 4. Class B placement posterior versus identity to nearest human keratinEach point is one class B (alpha-keratin) fragment, n = 33. Linear fit: slope −0.0098 per percentage point, R² = 0.88.Posterior placement probability0.000.100.200.300.400.506065707580859095Identity to nearest human keratin peptide (%)Nine fragments at 82–91%: two…Class B fragments linear fit: y = -0.00981x + 0.962, R² = 0.88Class B fragments Identity to nearest human keratin peptide (%): 93 Posterior placement probability: 0.06Class B fragments Identity to nearest human keratin peptide (%): 91 Posterior placement probability: 0.08Class B fragments Identity to nearest human keratin peptide (%): 90 Posterior placement probability: 0.05Class B fragments Identity to nearest human keratin peptide (%): 89 Posterior placement probability: 0.11Class B fragments Identity to nearest human keratin peptide (%): 88 Posterior placement probability: 0.09Class B fragments Identity to nearest human keratin peptide (%): 88 Posterior placement probability: 0.14Class B fragments Identity to nearest human keratin peptide (%): 87 Posterior placement probability: 0.07Class B fragments Identity to nearest human keratin peptide (%): 86 Posterior placement probability: 0.12Class B fragments Identity to nearest human keratin peptide (%): 85 Posterior placement probability: 0.16Class B fragments Identity to nearest human keratin peptide (%): 84 Posterior placement probability: 0.10Class B fragments Identity to nearest human keratin peptide (%): 83 Posterior placement probability: 0.18Class B fragments Identity to nearest human keratin peptide (%): 82 Posterior placement probability: 0.13Class B fragments Identity to nearest human keratin peptide (%): 81 Posterior placement probability: 0.21Class B fragments Identity to nearest human keratin peptide (%): 80 Posterior placement probability: 0.17Class B fragments Identity to nearest human keratin peptide (%): 79 Posterior placement probability: 0.15Class B fragments Identity to nearest human keratin peptide (%): 78 Posterior placement probability: 0.24Class B fragments Identity to nearest human keratin peptide (%): 77 Posterior placement probability: 0.19Class B fragments Identity to nearest human keratin peptide (%): 76 Posterior placement probability: 0.22Class B fragments Identity to nearest human keratin peptide (%): 75 Posterior placement probability: 0.18Class B fragments Identity to nearest human keratin peptide (%): 74 Posterior placement probability: 0.27Class B fragments Identity to nearest human keratin peptide (%): 73 Posterior placement probability: 0.20Class B fragments Identity to nearest human keratin peptide (%): 72 Posterior placement probability: 0.29Class B fragments Identity to nearest human keratin peptide (%): 71 Posterior placement probability: 0.25Class B fragments Identity to nearest human keratin peptide (%): 70 Posterior placement probability: 0.23Class B fragments Identity to nearest human keratin peptide (%): 69 Posterior placement probability: 0.31Class B fragments Identity to nearest human keratin peptide (%): 68 Posterior placement probability: 0.28Class B fragments Identity to nearest human keratin peptide (%): 67 Posterior placement probability: 0.34Class B fragments Identity to nearest human keratin peptide (%): 66 Posterior placement probability: 0.26Class B fragments Identity to nearest human keratin peptide (%): 65 Posterior placement probability: 0.37Class B fragments Identity to nearest human keratin peptide (%): 64 Posterior placement probability: 0.30Class B fragments Identity to nearest human keratin peptide (%): 63 Posterior placement probability: 0.41Class B fragments Identity to nearest human keratin peptide (%): 63 Posterior placement probability: 0.33Class B fragments Identity to nearest human keratin peptide (%): 62 Posterior placement probability: 0.36Figure 4. Class B placement posterior versus identity to nearest human keratinEach point is one class B (alpha-keratin) fragment, n = 33. Linear fit: slope −0.0098 per percentage point, R² = 0.88.Posterior placement probability0.000.100.200.300.400.5060708090Identity to nearest human keratin peptide (%)Nine fragments at82–91%:…Class B fragments linear fit: y = -0.00981x + 0.962, R² = 0.88Class B fragments Identity to nearest human keratin peptide (%): 93 Posterior placement probability: 0.06Class B fragments Identity to nearest human keratin peptide (%): 91 Posterior placement probability: 0.08Class B fragments Identity to nearest human keratin peptide (%): 90 Posterior placement probability: 0.05Class B fragments Identity to nearest human keratin peptide (%): 89 Posterior placement probability: 0.11Class B fragments Identity to nearest human keratin peptide (%): 88 Posterior placement probability: 0.09Class B fragments Identity to nearest human keratin peptide (%): 88 Posterior placement probability: 0.14Class B fragments Identity to nearest human keratin peptide (%): 87 Posterior placement probability: 0.07Class B fragments Identity to nearest human keratin peptide (%): 86 Posterior placement probability: 0.12Class B fragments Identity to nearest human keratin peptide (%): 85 Posterior placement probability: 0.16Class B fragments Identity to nearest human keratin peptide (%): 84 Posterior placement probability: 0.10Class B fragments Identity to nearest human keratin peptide (%): 83 Posterior placement probability: 0.18Class B fragments Identity to nearest human keratin peptide (%): 82 Posterior placement probability: 0.13Class B fragments Identity to nearest human keratin peptide (%): 81 Posterior placement probability: 0.21Class B fragments Identity to nearest human keratin peptide (%): 80 Posterior placement probability: 0.17Class B fragments Identity to nearest human keratin peptide (%): 79 Posterior placement probability: 0.15Class B fragments Identity to nearest human keratin peptide (%): 78 Posterior placement probability: 0.24Class B fragments Identity to nearest human keratin peptide (%): 77 Posterior placement probability: 0.19Class B fragments Identity to nearest human keratin peptide (%): 76 Posterior placement probability: 0.22Class B fragments Identity to nearest human keratin peptide (%): 75 Posterior placement probability: 0.18Class B fragments Identity to nearest human keratin peptide (%): 74 Posterior placement probability: 0.27Class B fragments Identity to nearest human keratin peptide (%): 73 Posterior placement probability: 0.20Class B fragments Identity to nearest human keratin peptide (%): 72 Posterior placement probability: 0.29Class B fragments Identity to nearest human keratin peptide (%): 71 Posterior placement probability: 0.25Class B fragments Identity to nearest human keratin peptide (%): 70 Posterior placement probability: 0.23Class B fragments Identity to nearest human keratin peptide (%): 69 Posterior placement probability: 0.31Class B fragments Identity to nearest human keratin peptide (%): 68 Posterior placement probability: 0.28Class B fragments Identity to nearest human keratin peptide (%): 67 Posterior placement probability: 0.34Class B fragments Identity to nearest human keratin peptide (%): 66 Posterior placement probability: 0.26Class B fragments Identity to nearest human keratin peptide (%): 65 Posterior placement probability: 0.37Class B fragments Identity to nearest human keratin peptide (%): 64 Posterior placement probability: 0.30Class B fragments Identity to nearest human keratin peptide (%): 63 Posterior placement probability: 0.41Class B fragments Identity to nearest human keratin peptide (%): 63 Posterior placement probability: 0.33Class B fragments Identity to nearest human keratin peptide (%): 62 Posterior placement probability: 0.36
Figure 4.Class B placement posterior versus identity to nearest human keratin. Each point is one class B (alpha-keratin) fragment, n = 33. Linear fit: slope −0.0098 per percentage point, R² = 0.88.Source: JM Shenzhen Research Center, build manifest v2.3.
Source data
Identity to nearest human keratin peptide (%)Posterior placement probability
930.06
910.08
900.05
890.11
880.09
880.14
870.07
860.12
850.16
840.1
830.18
820.13
810.21
800.17
790.15
780.24
770.19
760.22
750.18
740.27
730.2
720.29
710.25
700.23
690.31
680.28
670.34
660.26
650.37
640.3
630.41
630.33
620.36

4.3 Controls

Every sheath extraction batch contained human keratin in both procedural blanks, as in every keratin- handling proteomics laboratory (Table 3). No blank contained a peptide matching any archosaur CBP. In one blank (batch SX-2020-04), two PSMs matched a class B fragment that was also identified in specimen JM-SP-0044, extracted in batch SX-2020-03. Under the standing rule, which applies to blanks from the same batch, the fragment was retained and is flagged.

Table 3. Procedural blanks for sheath extraction batches.
Batch Specimen Blanks Human keratin PSMs (mean per blank) Archosaur CBP PSMs Archosaur alpha-keratin PSMs Porcine trypsin PSMs
SX-2020-01 JM-SP-0043 2 214 0 0 41
SX-2020-02 JM-SP-0043 2 187 0 0 38
SX-2020-03 JM-SP-0044 2 236 0 0 44
SX-2020-04 JM-SP-0044 2 198 0 2 36
SX-2020-05 JM-SP-0043, JM-SP-0044 2 171 0 0 40

4.4 Reference sensitivity

Placement of non-KAP fragments improves steadily as reference taxa are added (Fig. 5). Placement of class A fragments is essentially flat above 19 taxa, and class B placement is flat throughout. The limitation is not reference sampling; it is the nature of the proteins.

Figure 5. Placement posterior versus reference-set size, by fragment classMedian posterior placement probability across 20 random reference subsets per size (v2.3 fragments). Band: 5th–95th percentile across subsets for non-KAP fragments.Median posterior placement probability0.000.200.400.600.801.00152025303540Extant reference taxaFull v2.3 reference setNon-KAP fragmentsClass A (CBP)Class B (alpha-keratin)Extant reference taxa: 12 Non-KAP fragments: 0.44 (0.39–0.49) Class A (CBP): 0.31 Class B (alpha-keratin): 0.20Extant reference taxa: 19 Non-KAP fragments: 0.52 (0.48–0.56) Class A (CBP): 0.33 Class B (alpha-keratin): 0.19Extant reference taxa: 26 Non-KAP fragments: 0.59 (0.56–0.62) Class A (CBP): 0.34 Class B (alpha-keratin): 0.18Extant reference taxa: 34 Non-KAP fragments: 0.64 (0.62–0.66) Class A (CBP): 0.34 Class B (alpha-keratin): 0.18Extant reference taxa: 41 Non-KAP fragments: 0.67 (0.67–0.67) Class A (CBP): 0.34 Class B (alpha-keratin): 0.20Figure 5. Placement posterior versus reference-set size, by fragment classMedian posterior placement probability across 20 random reference subsets per size (v2.3 fragments). Band: 5th–95th percentile across subsets for non-KAP fragments.Median posterior placement probability0.000.200.400.600.801.00152025303540Extant reference taxaFull v2.3 referencesetExtant reference taxa: 12 Non-KAP fragments: 0.44 (0.39–0.49) Class A (CBP): 0.31 Class B (alpha-keratin): 0.20Extant reference taxa: 19 Non-KAP fragments: 0.52 (0.48–0.56) Class A (CBP): 0.33 Class B (alpha-keratin): 0.19Extant reference taxa: 26 Non-KAP fragments: 0.59 (0.56–0.62) Class A (CBP): 0.34 Class B (alpha-keratin): 0.18Extant reference taxa: 34 Non-KAP fragments: 0.64 (0.62–0.66) Class A (CBP): 0.34 Class B (alpha-keratin): 0.18Extant reference taxa: 41 Non-KAP fragments: 0.67 (0.67–0.67) Class A (CBP): 0.34 Class B (alpha-keratin): 0.20
Figure 5.Placement posterior versus reference-set size, by fragment class. Median posterior placement probability across 20 random reference subsets per size (v2.3 fragments). Band: 5th–95th percentile across subsets for non-KAP fragments.Source: JM Shenzhen Research Center, reference-subsampling run RS-2020-11.
Source data
Extant reference taxaNon-KAP fragmentsNon-KAP fragments intervalClass A (CBP)Class B (alpha-keratin)
120.440.39–0.490.310.2
190.520.48–0.560.330.19
260.590.56–0.620.340.18
340.640.62–0.660.340.18
410.670.67–0.670.340.2

4.5 Disposition: flagged, not removed

We considered three options for the 87 KAPs: exclusion from the library, retention with exclusion from default builds, and retention in default builds with a flag. We chose the third. Removal of fragments because they place poorly would bias the library toward fragments that confirm the expected placement, and the v2.0 post-mortem (JM-TR-2019-004) demonstrated the consequences of fragments being excluded from builds without anyone noticing. A KAP with a diffuse placement contributes little to the joint likelihood, and its inclusion moves the library median by 0.01.

Each KAP carries the flag LOWCONF_KAP in its row metadata. Flagged entries are included in default builds at full weight. A build profile excluding flagged entries (exclude_flags: [LOWCONF_KAP]) is available and is not the default. No reconstruction target released to cell-line development since v2.3 release candidate 1 has been built with that profile. The disposition was reviewed by the Ethics & Bioresponsibility Council on 26 January 2021: reviewed; no concerns raised.

4.6 Chain-of-custody schema

The schema defines 25 fields in nine groups (Table 4). Twenty-three are mandatory. A fossil-derived fragment is linked through its row metadata to a specimen record, an extraction batch and one or more raw files identified by SHA-256 hash; a build fails if any link cannot be resolved. A missing fact is recorded as unknown and is not a build failure; an empty mandatory field is.

Table 4. Chain-of-custody metadata schema (v1.0, effective 2020-07-01).
Group Field Type Required Description
Identity specimen_id string Yes Accession identifier, JM-SP-NNNN
Identity element enum Yes Skeletal element (tooth, ungual, long bone, rib, other)
Identity attribution_at_acquisition string Yes Taxonomic attribution as stated by the source
Identity attribution_current string Yes Current internal attribution and basis
Identity provenance_tier enum Yes A, B or C, assigned at intake under SOP-SP-003 (JM-TR-2016-003)
Acquisition acquisition_channel enum Yes institutional_loan, field_collection, commercial_purchase, donation
Acquisition source_party string Yes Supplying person or organisation (role only in public outputs)
Acquisition acquisition_date date | unknown Yes Date of transfer to JM or to the immediate prior holder
Locality locality_stated string Yes Locality as stated by the source
Locality locality_verified boolean Yes Whether locality is supported by independent evidence
Locality stratigraphic_unit string | unknown Yes Formation or group
Documentation export_documentation_ref string | none_on_file Yes Reference to export documentation
Documentation import_documentation_ref string | none_on_file Yes Reference to import documentation
Custody custody_events[] list Yes Timestamped transfers: custodian, location, action, seal ID
Custody custody_gaps[] list Yes Periods with no recorded custodian, with start, end and description
Custody custody_status enum Yes COMPLETE, GAPS_DOCUMENTED, INCOMPLETE
Custody provenance_backfilled boolean Yes Record compiled after accession from historical documents
Storage storage_location string Yes Controlled-storage location and access class
Preparation consolidant_history string | unknown Yes Consolidants, adhesives and fillers known or observed
Preparation preparation_history string | unknown Yes Mechanical or chemical preparation before receipt
Preparation subsampling_events[] list Yes Each removal of material, with mass and purpose
Analytical linkage extraction_batch_ids[] list Yes Batches, each linked to its blanks
Analytical linkage raw_file_sha256[] list Yes Hashes of all raw LC-MS/MS files
Review ethics_review_ref string No Ethics & Bioresponsibility Council reference
Review notes text No Free text

4.7 Backfill results

Backfill was completed for all nine pre-schema specimens on 14 December 2020. Completeness by field group is shown in Fig. 6 and custody status in Table 5. Six specimens carry documented custody gaps. Seven of 14 have a verified locality. Export documentation is recorded as none_on_file for four specimens (JM-SP-0001, JM-SP-0011, JM-SP-0032 and JM-SP-0035). For the three Tier B specimens among them, an export permit was sighted at intake and recorded in the paper register (JM-TR-2016-003); the permits themselves could not be located during backfill. Tier is unchanged.

Figure 6. Chain-of-custody record completeness by specimen and field groupFraction of fields in each group populated with a value other than unknown or noneonfile. JM-SP-0001 to JM-SP-0035 were backfilled; JM-SP-0041 to JM-SP-0045 were recorded at accession.Field groupIdentityAcquisitionLocalityDocumentationCustodyStoragePreparationAnalytical linkageJM-SP-0001JM-SP-0001 × Identity: 11.00JM-SP-0001 × Acquisition: 0.670.67JM-SP-0001 × Locality: 0.330.33JM-SP-0001 × Documentation: 00.00JM-SP-0001 × Custody: 0.50.50JM-SP-0001 × Storage: 11.00JM-SP-0001 × Preparation: 0.330.33JM-SP-0001 × Analytical linkage: 11.00JM-SP-0011JM-SP-0011 × Identity: 11.00JM-SP-0011 × Acquisition: 11.00JM-SP-0011 × Locality: 0.330.33JM-SP-0011 × Documentation: 00.00JM-SP-0011 × Custody: 0.750.75JM-SP-0011 × Storage: 11.00JM-SP-0011 × Preparation: 0.670.67JM-SP-0011 × Analytical linkage: 11.00JM-SP-0014JM-SP-0014 × Identity: 11.00JM-SP-0014 × Acquisition: 11.00JM-SP-0014 × Locality: 11.00JM-SP-0014 × Documentation: 11.00JM-SP-0014 × Custody: 11.00JM-SP-0014 × Storage: 11.00JM-SP-0014 × Preparation: 11.00JM-SP-0014 × Analytical linkage: 11.00JM-SP-0019JM-SP-0019 × Identity: 11.00JM-SP-0019 × Acquisition: 11.00JM-SP-0019 × Locality: 11.00JM-SP-0019 × Documentation: 11.00JM-SP-0019 × Custody: 11.00JM-SP-0019 × Storage: 11.00JM-SP-0019 × Preparation: 0.670.67JM-SP-0019 × Analytical linkage: 11.00JM-SP-0031JM-SP-0031 × Identity: 11.00JM-SP-0031 × Acquisition: 11.00JM-SP-0031 × Locality: 0.670.67JM-SP-0031 × Documentation: 11.00JM-SP-0031 × Custody: 11.00JM-SP-0031 × Storage: 11.00JM-SP-0031 × Preparation: 11.00JM-SP-0031 × Analytical linkage: 11.00JM-SP-0032JM-SP-0032 × Identity: 11.00JM-SP-0032 × Acquisition: 11.00JM-SP-0032 × Locality: 0.330.33JM-SP-0032 × Documentation: 00.00JM-SP-0032 × Custody: 0.750.75JM-SP-0032 × Storage: 11.00JM-SP-0032 × Preparation: 0.670.67JM-SP-0032 × Analytical linkage: 11.00JM-SP-0033JM-SP-0033 × Identity: 11.00JM-SP-0033 × Acquisition: 11.00JM-SP-0033 × Locality: 11.00JM-SP-0033 × Documentation: 11.00JM-SP-0033 × Custody: 11.00JM-SP-0033 × Storage: 11.00JM-SP-0033 × Preparation: 11.00JM-SP-0033 × Analytical linkage: 11.00JM-SP-0034JM-SP-0034 × Identity: 11.00JM-SP-0034 × Acquisition: 11.00JM-SP-0034 × Locality: 11.00JM-SP-0034 × Documentation: 11.00JM-SP-0034 × Custody: 11.00JM-SP-0034 × Storage: 11.00JM-SP-0034 × Preparation: 11.00JM-SP-0034 × Analytical linkage: 11.00JM-SP-0035JM-SP-0035 × Identity: 11.00JM-SP-0035 × Acquisition: 11.00JM-SP-0035 × Locality: 0.330.33JM-SP-0035 × Documentation: 00.00JM-SP-0035 × Custody: 0.750.75JM-SP-0035 × Storage: 11.00JM-SP-0035 × Preparation: 0.670.67JM-SP-0035 × Analytical linkage: 11.00JM-SP-0041JM-SP-0041 × Identity: 11.00JM-SP-0041 × Acquisition: 11.00JM-SP-0041 × Locality: 11.00JM-SP-0041 × Documentation: 11.00JM-SP-0041 × Custody: 11.00JM-SP-0041 × Storage: 11.00JM-SP-0041 × Preparation: 11.00JM-SP-0041 × Analytical linkage: 11.00JM-SP-0042JM-SP-0042 × Identity: 11.00JM-SP-0042 × Acquisition: 11.00JM-SP-0042 × Locality: 11.00JM-SP-0042 × Documentation: 11.00JM-SP-0042 × Custody: 11.00JM-SP-0042 × Storage: 11.00JM-SP-0042 × Preparation: 11.00JM-SP-0042 × Analytical linkage: 11.00JM-SP-0043JM-SP-0043 × Identity: 11.00JM-SP-0043 × Acquisition: 11.00JM-SP-0043 × Locality: 0.670.67JM-SP-0043 × Documentation: 11.00JM-SP-0043 × Custody: 0.750.75JM-SP-0043 × Storage: 11.00JM-SP-0043 × Preparation: 0.670.67JM-SP-0043 × Analytical linkage: 11.00JM-SP-0044JM-SP-0044 × Identity: 11.00JM-SP-0044 × Acquisition: 11.00JM-SP-0044 × Locality: 0.670.67JM-SP-0044 × Documentation: 11.00JM-SP-0044 × Custody: 0.750.75JM-SP-0044 × Storage: 11.00JM-SP-0044 × Preparation: 0.670.67JM-SP-0044 × Analytical linkage: 11.00JM-SP-0045JM-SP-0045 × Identity: 11.00JM-SP-0045 × Acquisition: 11.00JM-SP-0045 × Locality: 11.00JM-SP-0045 × Documentation: 11.00JM-SP-0045 × Custody: 11.00JM-SP-0045 × Storage: 11.00JM-SP-0045 × Preparation: 11.00JM-SP-0045 × Analytical linkage: 11.00Figure 6. Chain-of-custody record completeness by specimen and field groupFraction of fields in each group populated with a value other than unknown or noneonfile. JM-SP-0001 to JM-SP-0035 were backfilled; JM-SP-0041 to JM-SP-0045 were recorded at accession.Field groupIdentityAcquisitionLocalityDocumentationCustodyStoragePreparationAnalytical linkageJM-SP-0001JM-SP-0001 × Identity: 1JM-SP-0001 × Acquisition: 0.67JM-SP-0001 × Locality: 0.33JM-SP-0001 × Documentation: 0JM-SP-0001 × Custody: 0.5JM-SP-0001 × Storage: 1JM-SP-0001 × Preparation: 0.33JM-SP-0001 × Analytical linkage: 1JM-SP-0011JM-SP-0011 × Identity: 1JM-SP-0011 × Acquisition: 1JM-SP-0011 × Locality: 0.33JM-SP-0011 × Documentation: 0JM-SP-0011 × Custody: 0.75JM-SP-0011 × Storage: 1JM-SP-0011 × Preparation: 0.67JM-SP-0011 × Analytical linkage: 1JM-SP-0014JM-SP-0014 × Identity: 1JM-SP-0014 × Acquisition: 1JM-SP-0014 × Locality: 1JM-SP-0014 × Documentation: 1JM-SP-0014 × Custody: 1JM-SP-0014 × Storage: 1JM-SP-0014 × Preparation: 1JM-SP-0014 × Analytical linkage: 1JM-SP-0019JM-SP-0019 × Identity: 1JM-SP-0019 × Acquisition: 1JM-SP-0019 × Locality: 1JM-SP-0019 × Documentation: 1JM-SP-0019 × Custody: 1JM-SP-0019 × Storage: 1JM-SP-0019 × Preparation: 0.67JM-SP-0019 × Analytical linkage: 1JM-SP-0031JM-SP-0031 × Identity: 1JM-SP-0031 × Acquisition: 1JM-SP-0031 × Locality: 0.67JM-SP-0031 × Documentation: 1JM-SP-0031 × Custody: 1JM-SP-0031 × Storage: 1JM-SP-0031 × Preparation: 1JM-SP-0031 × Analytical linkage: 1JM-SP-0032JM-SP-0032 × Identity: 1JM-SP-0032 × Acquisition: 1JM-SP-0032 × Locality: 0.33JM-SP-0032 × Documentation: 0JM-SP-0032 × Custody: 0.75JM-SP-0032 × Storage: 1JM-SP-0032 × Preparation: 0.67JM-SP-0032 × Analytical linkage: 1JM-SP-0033JM-SP-0033 × Identity: 1JM-SP-0033 × Acquisition: 1JM-SP-0033 × Locality: 1JM-SP-0033 × Documentation: 1JM-SP-0033 × Custody: 1JM-SP-0033 × Storage: 1JM-SP-0033 × Preparation: 1JM-SP-0033 × Analytical linkage: 1JM-SP-0034JM-SP-0034 × Identity: 1JM-SP-0034 × Acquisition: 1JM-SP-0034 × Locality: 1JM-SP-0034 × Documentation: 1JM-SP-0034 × Custody: 1JM-SP-0034 × Storage: 1JM-SP-0034 × Preparation: 1JM-SP-0034 × Analytical linkage: 1JM-SP-0035JM-SP-0035 × Identity: 1JM-SP-0035 × Acquisition: 1JM-SP-0035 × Locality: 0.33JM-SP-0035 × Documentation: 0JM-SP-0035 × Custody: 0.75JM-SP-0035 × Storage: 1JM-SP-0035 × Preparation: 0.67JM-SP-0035 × Analytical linkage: 1JM-SP-0041JM-SP-0041 × Identity: 1JM-SP-0041 × Acquisition: 1JM-SP-0041 × Locality: 1JM-SP-0041 × Documentation: 1JM-SP-0041 × Custody: 1JM-SP-0041 × Storage: 1JM-SP-0041 × Preparation: 1JM-SP-0041 × Analytical linkage: 1JM-SP-0042JM-SP-0042 × Identity: 1JM-SP-0042 × Acquisition: 1JM-SP-0042 × Locality: 1JM-SP-0042 × Documentation: 1JM-SP-0042 × Custody: 1JM-SP-0042 × Storage: 1JM-SP-0042 × Preparation: 1JM-SP-0042 × Analytical linkage: 1JM-SP-0043JM-SP-0043 × Identity: 1JM-SP-0043 × Acquisition: 1JM-SP-0043 × Locality: 0.67JM-SP-0043 × Documentation: 1JM-SP-0043 × Custody: 0.75JM-SP-0043 × Storage: 1JM-SP-0043 × Preparation: 0.67JM-SP-0043 × Analytical linkage: 1JM-SP-0044JM-SP-0044 × Identity: 1JM-SP-0044 × Acquisition: 1JM-SP-0044 × Locality: 0.67JM-SP-0044 × Documentation: 1JM-SP-0044 × Custody: 0.75JM-SP-0044 × Storage: 1JM-SP-0044 × Preparation: 0.67JM-SP-0044 × Analytical linkage: 1JM-SP-0045JM-SP-0045 × Identity: 1JM-SP-0045 × Acquisition: 1JM-SP-0045 × Locality: 1JM-SP-0045 × Documentation: 1JM-SP-0045 × Custody: 1JM-SP-0045 × Storage: 1JM-SP-0045 × Preparation: 1JM-SP-0045 × Analytical linkage: 1
Figure 6.Chain-of-custody record completeness by specimen and field group. Fraction of fields in each group populated with a value other than unknown or none_on_file. JM-SP-0001 to JM-SP-0035 were backfilled; JM-SP-0041 to JM-SP-0045 were recorded at accession.Source: Specimen custody register, snapshot 2021-01-29.
Source data
SpecimenIdentityAcquisitionLocalityDocumentationCustodyStoragePreparationAnalytical linkage
JM-SP-000110.670.3300.510.331
JM-SP-0011110.3300.7510.671
JM-SP-001411111111
JM-SP-00191111110.671
JM-SP-0031110.6711111
JM-SP-0032110.3300.7510.671
JM-SP-003311111111
JM-SP-003411111111
JM-SP-0035110.3300.7510.671
JM-SP-004111111111
JM-SP-004211111111
JM-SP-0043110.6710.7510.671
JM-SP-0044110.6710.7510.671
JM-SP-004511111111

Table 5. Custody status by specimen, Library v2.3.
Specimen Element Channel Tier Custody status Backfilled Locality verified Fragments in v2.3
JM-SP-0001 Tooth donation C GAPS_DOCUMENTED Yes No 452
JM-SP-0011 Tooth commercial_purchase B GAPS_DOCUMENTED Yes No 158
JM-SP-0014 Rib institutional_loan A COMPLETE Yes Yes 197
JM-SP-0019 Tooth commercial_purchase A COMPLETE Yes Yes 94
JM-SP-0031 Tooth commercial_purchase B COMPLETE Yes No 92
JM-SP-0032 Tooth commercial_purchase B GAPS_DOCUMENTED Yes No 64
JM-SP-0033 Long bone institutional_loan A COMPLETE Yes Yes 98
JM-SP-0034 Long bone institutional_loan A COMPLETE Yes Yes 71
JM-SP-0035 Rib commercial_purchase B GAPS_DOCUMENTED Yes No 58
JM-SP-0041 Tooth institutional_loan A COMPLETE No Yes 71
JM-SP-0042 Long bone institutional_loan A COMPLETE No Yes 98
JM-SP-0043 Ungual commercial_purchase B GAPS_DOCUMENTED No No 52
JM-SP-0044 Ungual commercial_purchase B GAPS_DOCUMENTED No No 41
JM-SP-0045 Long bone institutional_loan A COMPLETE No Yes 64
Total 8 complete, 6 with gaps 9 7 1,610

The record for the founding specimen illustrates the backfill. JM-SP-0001 is recorded with attribution_at_acquisition "Carcharodontosaurus (vendor)", attribution_current "Carcharodontosauridae indet. (JM-TR-2016-003)", provenance_tier C, acquisition_channel donation (from a private collector who had purchased it at the Rissani Sunday market, Tafilalt, before 2011), acquisition_date unknown, locality_stated "Kem Kem Group, southeastern Morocco", locality_verified false, export_documentation_ref none_on_file, and consolidant_history "cyanoacrylate at mid-crown fracture; adhesive and pigmented mineral filler bonding a root from a different tooth at the crown base; both observed at receipt" (JM-TR-2013-001). It carries two custody gaps: the period before November 2011, for which no custodian, location or duration is recorded beyond a residential display cabinet; and postal transit from the collector to the Tsinghua University laboratory in November 2011 (11 days, unbagged). The specimen's analytical linkage is complete for all 87 raw files.

5. Discussion

The keratin-associated peptides present a clean illustration of a general problem. The class of protein that is most robust to contamination criticism (CBPs, absent from mammals) is the class that is least informative for placement, because of its gene-family structure; and the class that would be informative if it placed well (alpha-keratins) is the class most exposed to contamination. Neither improves with more reference data (Fig. 5). Recent work recovering phylogenetically informative enamel proteomes from Pleistocene and Pliocene material (Cappellini et al. 2019; Welker et al. 2019) reinforces that tooth-derived enamel proteins, rather than keratins, are the more promising route to placement, and enamel remains a priority for specimen acquisition.

The retention of KAPs in default builds reflects a principle adopted after JM-TR-2019-004: the library should not exclude fragments for performing badly. We accept that this makes the headline median 0.01 lower than it would otherwise be. We also note, for completeness, that the nine class B fragments differing from human keratin only at deamidation-consistent positions are retained by the same rule that, under an alternative and equally defensible treatment of deamidation, would exclude them. The standing rule has not been changed in this release. Treatment of diagenetic modification in placement and screening is scheduled for a future release.

The chain-of-custody schema converts specimen documentation from narrative to structured data. Its principal effect so far has been to make the gaps countable. Six of 14 specimens, contributing 51.2% of library fragments, carry documented gaps. The schema does not, by design, make custody status an inclusion criterion: excluding the six specimens would remove the founding specimen and every keratin-associated fragment, and would leave a library of 785 fragments. Custody status is reported in every build manifest so that downstream users can apply their own criteria.

6. Limitations and known issues

  • Certain keratin-associated peptides produce low-confidence placement (median posterior 0.29). They are flagged and included in default builds.
  • The contamination rule counts deamidation-consistent differences as substitutions. Nine class B fragments pass the rule only on this basis.
  • Four class B fragments excluded by the contamination rule have been restored by curator override.
  • One class B fragment identified in JM-SP-0044 was also detected in a procedural blank from an adjacent batch. The standing rule applies to same-batch blanks only.
  • Locality is unverified for seven of 14 specimens. For specimens obtained through commercial channels, locality depends on the vendor's statement.
  • Export documentation is not on file for four specimens. These records are documented as pre-dating current procedure. For three of them the permit was sighted at intake and has since not been located.
  • Backfilled records rely on recollection for some pre-2014 events. Recollection is recorded as a source type and is not independently verifiable.
  • Consolidants observed on JM-SP-0001, JM-SP-0043 and JM-SP-0044 were characterised at receipt; consolidants applied before receipt and since removed cannot be characterised.
  • Override records do not have a mandatory rationale field or an expiry. The 1,207 legacy records created at the v2.0 migration (41 converted v1.x override flags and 1,166 unflagged manual edits; JM-TR-2019-004, CA-6) remain attributed to the migration service account jm-curate.

7. Changes in Library v2.3

  • Known issue: certain keratin-associated peptides produce low-confidence placement. Flagged, not removed.
  • Added chain-of-custody metadata fields for all fossil-derived entries.
  • Fixed: specimen identifiers for two v1.4 entries had been transposed. Corrected. Placement is unaffected, because both entries derive from the same specimen batch.
  • --no-bracket-filter disabled ahead of its scheduled removal.
  • Seven extant reference taxa added (34 → 41); 326 theropod-assigned fossil-derived entries added (1,284 → 1,610).

The known issue is described in Sections 4.2–4.5; the custody fields in Sections 4.6–4.7 and Table 4. The transposed identifiers were found during backfill (Section 4.7): one fragment recorded against JM-SP-0019 was from JM-SP-0014, and one recorded against JM-SP-0014 was from JM-SP-0019. The two specimens were extracted in the same batch (JM-TR-2016-003, Section 4.3). The placement_flag field listed in the release notes carries the value LOWCONF_KAP for all 87 keratin-associated peptides.

8. Data availability

Library contents are confidential. Library access is restricted to authorized research staff and is not available to third parties. Specimen custody records are held in the specimen custody register and are available to auditors and regulators on request under confidentiality. Aggregate statistics are derived from build manifest v2.3. Questions about this report may be directed to regulatory@jurassicmeat.com.

9. Acknowledgements

We thank the staff of the clean-room suite for the sheath extraction work, the Computational Genomics group for the reference-subsampling runs, and the lending institutions for specimens JM-SP-0033, JM-SP-0034, JM-SP-0041, JM-SP-0042 and JM-SP-0045. We thank a dealer in the Rissani Sunday market for assistance with specimen provenance. The disposition of keratin-associated peptides and the custody schema were reviewed by the Ethics & Bioresponsibility Council: reviewed; no concerns raised. This work was funded internally. All authors are employees of Jurassic Meat and may hold equity.

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. Cappellini E, et al. (2019). Early Pleistocene enamel proteome from Dmanisi resolves Stephanorhinus phylogeny. Nature 574:103–107.
  4. 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.
  5. Demarchi B, et al. (2016). Protein sequences bound to mineral surfaces persist into deep time. eLife 5:e17092.
  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. Schweitzer MH, et al. (2007). Analyses of soft tissue from Tyrannosaurus rex suggest the presence of protein. Science 316:277–280.
  8. Welker F, et al. (2019). Enamel proteome shows that Gigantopithecus was an early diverging pongine. Nature 576:262–265.
  9. JM-TR-2013-001. Endogenous peptide fragments recovered from a Kem Kem Group theropod tooth. Jurassic Meat internal technical report.
  10. JM-TR-2016-003. Library v1.4: fragments from additional specimens with documented provenance. Jurassic Meat internal technical report.
  11. JM-TR-2019-004. Library v2.0: a versioned multiple-sequence-alignment architecture for fragmentary ancient peptides. Jurassic Meat internal technical report.