申请配额 联系我们

科学技术报告

JM-TR-2014-002 技术报告 文库 v1.0

Peptide Library v1.0: consolidation of the JM-SP-0001 tooth-derived fragments, extant archosaur reference import, and a manual curation procedure

Dr. Yuting Zhou 1,* Dr. Hao Lin 1 Dr. Lei Chen 1

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

目录
  1. 摘要
  2. 1. Summary
  3. 2. Background
  4. 3. Materials and methods
  5. 4. Results
  6. 5. Discussion
  7. 6. Limitations and known issues
  8. 7. Changes in Library v1.0
  9. 8. Data availability
  10. 9. Acknowledgements
  11. 10. References
  12. Supplementary Table S1. Fragment entry fields

摘要

Library v1.0 is the first structured release of the Jurassic Meat peptide library. It replaces 17 unversioned working files, maintained between 2011 and 2013 for the analysis of a single Kem Kem Group theropod tooth (specimen JM-SP-0001; JM-TR-2013-001), with a single versioned reference set in which every fossil-derived fragment carries a stable identifier, its authentication evidence and its curation history. Reconciliation of the working files identified 63 annotation conflicts, all resolved. Archived raw data from the 2011–2012 extraction campaigns were re-searched against an expanded reference database; no new specimen material was extracted. After merging three isobaric duplicates and admitting 21 newly identified peptides under unchanged authentication criteria, the library contains 214 theropod fragments covering 1,880 unique residues. The extant reference set was expanded from seven to twelve taxa (seven birds, three crocodilians, one turtle and one squamate) across seven proteins. Enamel matrix proteins, for which birds and turtles provide no functional comparison, are handled explicitly. All 214 entries were reviewed independently by two curators under a written procedure (SOP-PL-001); 37 manual override flags were applied. On the twelve-taxon reference tree, the median per-fragment placement posterior rises from 0.38 to 0.41, and the concatenated placement on the stem lineage of Aves rises from 0.91 to 0.93. Curation is manual only.

关键词palaeoproteomicsreference librarydata curationphylogenetic placementcollagen type Ienamel proteinsextant phylogenetic bracket

JM-TR-2014-002 · Public summary of internal report · Shenzhen, March 2014

1. Summary

  • Library v1.0 is the first structured release. It consolidates the tooth-derived theropod peptide fragments recovered in 2011–2012 into a single versioned reference set: 214 fragments, 1,880 unique residues, all from specimen JM-SP-0001.
  • No new extraction was performed. The 18 additional fragments relative to JM-TR-2013-001 come from re-searching archived raw files against an expanded reference database, net of three duplicate merges.
  • The extant reference set expands from 7 to 12 taxa, including the first squamate reference, which provides the only non-crocodilian comparison for enamel proteins.
  • Curation is manual. Every entry was reviewed by two curators under SOP-PL-001; 37 entries carry manual override flags.
  • Placement improves modestly. Median per-fragment posterior 0.41 (previously 0.38); concatenated stem-Aves posterior 0.93 (previously 0.91).

2. Background

JM-TR-2013-001 reported 196 authenticated peptides from an isolated theropod tooth crown from the Kem Kem Group of southeastern Morocco, and their preliminary placement on the stem lineage of Aves. That report concluded that the working files used to produce it were adequate for a single specimen and a single report, and inadequate for anything that would later depend on the sequences.

That conclusion has become more pressing. The company's research programme now has a stated downstream use for these peptides: as anchors and checkpoints for comparative reconstruction of extinct-species protein sequence by phylogenetic bracketing against living relatives (Witmer 1995). Any reconstruction that depends on a fossil peptide needs to be able to say which peptide, from which specimen, under which authentication evidence, in which version of the reference set. None of those questions could be answered reliably from the working files.

This report describes:

  1. the reconciliation of the working files into a single data model;
  2. a re-search of the archived 2011–2012 raw data against an expanded reference database;
  3. the import of extant avian, crocodilian and outgroup reference sequences;
  4. the manual curation procedure applied to every entry; and
  5. placement of the consolidated fragment set on the expanded reference tree.

All sequence information in the library is protein sequence. The library contains no nucleic acid sequence of fossil origin.

3. Materials and methods

3.1 Working-file inventory

All files used in the analyses reported in JM-TR-2013-001 were located on the two personal workstations and one shared network drive used during 2011–2013. Seventeen files were identified as containing peptide identifications, alignments or placements (Table 1). Each was copied to read-only storage with a checksum before any processing. Files were attributed to an author and date from file metadata and from the curators' records.

Table 1. Working files consolidated into Library v1.0.
File class Files Formats Rows / records Author attributed Notes
Identification spreadsheets 9 Spreadsheet (2 variants) 1,117 9 of 9 One per campaign and extract type, plus two merged summaries
Alignment files 5 Plain-text sequence alignment 205 4 of 5 COL1A1; COL1A2; enamel proteins (3); BGLAP and SPP1; one duplicate
Placement outputs 3 Tree files + tabular summaries 80 3 of 3 Directory per analysis run
Total 17 3 1,402 16 of 17

One alignment file could not be attributed to an author or date. Its contents were identical to another, attributed file, and it was retained.

3.2 Reconciliation rules

Records were matched across files by peptide sequence, with isoleucine and leucine treated as equivalent. Where records for the same sequence disagreed on any field, the conflict was logged and resolved by the following rules, in order:

  1. a value supported by the raw data on re-inspection was preferred to one that was not;
  2. a value recorded in the campaign-level spreadsheet was preferred to one recorded in a merged summary;
  3. otherwise, the conflict was referred to the curators under SOP-PL-001 (§3.5).

File modification timestamps were not used as evidence of precedence, as several files had been copied between machines with timestamps reset.

3.3 Re-search of archived raw data

The 41 raw files from Campaigns A–C (JM-TR-2013-001, Table 2), together with all 51 associated blank and control files, were re-searched with MaxQuant (Cox and Mann 2008) against the expanded reference database described in §3.4, concatenated with the unchanged contamination database (247 sequences) and reversed decoys. Search parameters, de novo procedure and the five authentication criteria were identical to those in JM-TR-2013-001, with one exception: the exclusion list derived from the 2011 Gallus method-development runs was applied to Meleagris identical matches as well, as the two are indistinguishable at most collagen positions.

3.4 Extant reference import

Reference sequences were imported for seven proteins (COL1A1, COL1A2, AMELX, ENAM, AMBN, BGLAP, SPP1) from twelve extant taxa (Table 2), drawn from public genome assemblies, gene models and protein databases as available in January 2014. Each imported sequence was checked against its gene model for signal peptide and propeptide annotation, and trimmed to the mature chain for collagens. Where multiple isoforms existed, the longest complete isoform was used.

Modern birds and turtles are edentulous. In both lineages the enamel matrix protein genes are absent from the assemblies or are present only as degraded remnants with frameshifts and premature stop codons. Such remnants were imported, labelled as pseudogene remnants, and excluded from placement scoring. They are retained because they constrain bracketing at conserved positions and because their removal would be harder to justify later than their retention.

Table 2. Extant reference taxa in Library v1.0. New in v1.0 relative to JM-TR-2013-001 marked †.
Taxon Common name Clade Collagens Enamel proteins BGLAP / SPP1
Gallus gallus Chicken Aves (Galloanserae) Complete Remnant Complete
Meleagris gallopavo † Turkey Aves (Galloanserae) Complete Remnant Partial
Anas platyrhynchos Mallard Aves (Galloanserae) Complete Absent Complete
Taeniopygia guttata Zebra finch Aves (Neoaves) Complete Remnant Complete
Columba livia † Rock pigeon Aves (Neoaves) Complete Absent Partial
Falco peregrinus † Peregrine falcon Aves (Neoaves) Partial Absent Partial
Struthio camelus Ostrich Aves (Palaeognathae) Partial Remnant Partial
Alligator mississippiensis American alligator Crocodylia Complete Complete Complete
Crocodylus porosus Saltwater crocodile Crocodylia Complete Partial Partial
Gavialis gangeticus † Gharial Crocodylia Partial Partial Partial
Chrysemys picta Painted turtle Testudines (outgroup) Complete Remnant Complete
Anolis carolinensis † Green anole Squamata (outgroup) Complete Complete Complete

3.5 Library data model

Each fossil-derived entry is a single peptide fragment, identified by a stable fragment ID of the form PL-F-000001, linked to a specimen ID (JM-SP-0001 for all entries in this release). Each extant reference entry is a single protein sequence for a single taxon. Fields recorded for each fragment entry are listed in Supplementary Table S1. The library is held as a set of tab-delimited tables and a per-protein alignment, under a version-control repository; each release is a tagged state of that repository. Version 1.0 is the first tag.

Peptides that failed authentication, or that were observed in a single campaign, are held in a separate non-production set (PL-U) with the same data model. PL-U is not used in placement or in any downstream reconstruction. At v1.0 it contains 548 entries: all 535 sequences excluded by the authentication cascade of JM-TR-2013-001 (83 of which appeared in the working files, the remainder recovered from archived search outputs), 2 re-search candidates that failed manual review, and 11 single-campaign sequences identified in the re-search.

3.6 Manual curation procedure (SOP-PL-001)

Every fragment entry was reviewed independently by two curators (Y.Z. and H.L.), each of whom recorded, without consulting the other: the protein assignment; the alignment coordinates; the resolution of any I/L position against the reference residue; the PTM annotation; and acceptance or rejection of the computed placement. Entries on which the two curators agreed in all fields were accepted. Discordant entries were discussed between the curators and, if still unresolved, adjudicated by the corresponding author's line manager (L.C.).

Where a curator's decision departed from the output of the automated search or placement step, the entry was marked with a manual override flag, a category, a free-text justification and the date, recorded under the shared curation account. Override flags are retained with the entry in all subsequent versions. They do not expire.

3.7 Phylogenetic placement

Placement followed JM-TR-2013-001 (§3.9), using the twelve-taxon reference topology shown in Fig. 5. Maximum-likelihood placement used a JTT+Γ model; Bayesian placement used two runs of four chains, 2×10⁶ generations, 25% burn-in. Pseudogene remnants were masked. Concatenated ML placement support was estimated from 1,000 bootstrap replicates.

3.8 Bracketing interface

Library v1.0 is the first version made available to the reconstruction team for phylogenetic bracketing. At v1.0, bracketing is performed manually from the curated per-protein alignment: each fragment is entered as a constraint by hand, and the curators confirmed that all 214 fragments were entered. An automated bracketing module, which will align fragments to the reference at build time and apply them as constraints without manual entry, is in development and is expected from v1.1.

4. Results

4.1 Reconciliation

The 1,402 records in the working files resolved to 279 distinct sequence records (Table 3). Of these, 196 corresponded to the authenticated set reported in JM-TR-2013-001 and 83 to unauthenticated or single-campaign sequences, which were moved to PL-U. Sixty-three conflicts were logged. Most were differences in PTM annotation between campaign-level and merged spreadsheets (31), followed by alignment offsets of one to three residues introduced when one alignment file was rebuilt by hand (18). All 63 were resolved; 49 by rules 1–2 and 14 by curator review.

Table 3. Reconciliation of working-file records.
Measure Count
Records across 17 working files 1,402
Distinct sequence records (I/L-equivalent) 279
— Matching JM-TR-2013-001 authenticated set 196
— Unauthenticated or single-campaign (moved to PL-U) 83
Conflicts logged 63
— PTM annotation 31
— Alignment offset (1–3 residues) 18
— Protein assignment (COL1A1 vs COL1A2) 9
— Campaign attribution 5
Resolved by reconciliation rules 49
Resolved by curator review 14
Unresolved 0

Three pairs within the authenticated set were found to be the same peptide recorded twice, differing only at an I/L position that had been resolved differently in two spreadsheets. Each pair was merged to a single entry, reducing the JM-TR-2013-001 set from 196 to 193.

4.2 Re-search against the expanded reference database

The re-search identified 23 candidate peptides not previously identified, all from collagen or enamel proteins, and all observed in at least two campaigns. Their earlier absence is attributable to sequence divergence from all seven original references sufficient to prevent identification at 1% FDR; most were identified via the Columba, Falco, Gavialis or Anolis sequences. Two failed manual spectral review. Twenty-one were admitted, giving 214 fragments in total (Fig. 1). No previously authenticated peptide was lost.

Figure 1. Authenticated fragments by protein: JM-TR-2013-001 and Library v1.0Fragment counts by protein in the authenticated set of JM-TR-2013-001 (n = 196) and in Library v1.0 (n = 214), after merging three I/L duplicates and admitting 21 fragments identified by re-search. All fragments derive from specimen JM-SP-0001.Fragments020406080100120140COL1A1COL1A1 JM-TR-2013-001: 118COL1A1 Library v1.0: 127COL1A2COL1A2 JM-TR-2013-001: 45COL1A2 Library v1.0: 51AMELXAMELX JM-TR-2013-001: 14AMELX Library v1.0: 15ENAMENAM JM-TR-2013-001: 7ENAM Library v1.0: 8AMBNAMBN JM-TR-2013-001: 5AMBN Library v1.0: 6BGLAP / SPP1BGLAP / SPP1 JM-TR-2013-001: 7BGLAP / SPP1 Library v1.0: 7ProteinFigure 1. Authenticated fragments by protein: JM-TR-2013-001 and Library v1.0Fragment counts by protein in the authenticated set of JM-TR-2013-001 (n = 196) and in Library v1.0 (n = 214), after merging three I/L duplicates and admitting 21 fragments identified by re-search. All fragments derive from specimen JM-SP-0001.Fragments020406080100120140COL1A1COL1A1 JM-TR-2013-001: 118COL1A1 Library v1.0: 127COL1A2COL1A2 JM-TR-2013-001: 45COL1A2 Library v1.0: 51AMELXAMELX JM-TR-2013-001: 14AMELX Library v1.0: 15ENAMENAM JM-TR-2013-001: 7ENAM Library v1.0: 8AMBNAMBN JM-TR-2013-001: 5AMBN Library v1.0: 6BGLAP/ SPP1BGLAP / SPP1 JM-TR-2013-001: 7BGLAP / SPP1 Library v1.0: 7Protein
Figure 1.Authenticated fragments by protein: JM-TR-2013-001 and Library v1.0. Fragment counts by protein in the authenticated set of JM-TR-2013-001 (n = 196) and in Library v1.0 (n = 214), after merging three I/L duplicates and admitting 21 fragments identified by re-search. All fragments derive from specimen JM-SP-0001.Source: JM Shenzhen, Peptide Library v1.0 (tag v1.0).
Source data
ProteinJM-TR-2013-001Library v1.0
COL1A1118127
COL1A24551
AMELX1415
ENAM78
AMBN56
BGLAP / SPP177

Unique residue coverage rose from 1,712 to 1,880. Of the 168 additional residues, 152 are contributed by newly admitted fragments and 16 by alignment corrections that extended the mapped length of existing fragments.

Table 4. Library v1.0 contents by protein.
Protein Fragments Unique residues Coverage of mature chain Diagnostic fragments Median length
COL1A1 127 816 77.2% 84 11
COL1A2 51 629 60.5% 36 12
AMELX 15 128 64.6% 13 10
ENAM 8 120 10.9% 8 14
AMBN 6 95 22.6% 6 13
BGLAP / SPP1 7 92 — 0 12
Total 214 1,880 147 11

BGLAP and SPP1 fragments remain excluded from placement, as in JM-TR-2013-001. Their unique-residue figure rises from 76 to 92 with no new fragments: all 16 alignment-correction residues (§4.1) fall in these two proteins, which share the alignment file that was rebuilt by hand (§4.1).

4.3 Reference coverage

The completeness of the extant reference set is shown in Fig. 2. Collagens are complete or near-complete for all taxa. Enamel matrix proteins are available in usable form only for the three crocodilians and the green anole; the anole is the first non-archosaur reference for enamel and partially offsets the crocodilian bias in enamel placement reported in JM-TR-2013-001.

Figure 2. Completeness of extant reference sequences by taxon and proteinFraction of the mature reference chain represented by usable sequence in Library v1.0. Pseudogene remnants (birds, turtle enamel proteins) are shown at 0.1 and are masked in placement. 0 = absent from available assemblies.ProteinCOL1A1COL1A2AMELXENAMAMBNBGLAPSPP1GallusGallus × COL1A1: 11.00Gallus × COL1A2: 11.00Gallus × AMELX: 0.10.10Gallus × ENAM: 0.10.10Gallus × AMBN: 0.10.10Gallus × BGLAP: 11.00Gallus × SPP1: 11.00MeleagrisMeleagris × COL1A1: 11.00Meleagris × COL1A2: 0.980.98Meleagris × AMELX: 0.10.10Meleagris × ENAM: 0.10.10Meleagris × AMBN: 00.00Meleagris × BGLAP: 11.00Meleagris × SPP1: 0.720.72AnasAnas × COL1A1: 11.00Anas × COL1A2: 11.00Anas × AMELX: 00.00Anas × ENAM: 00.00Anas × AMBN: 00.00Anas × BGLAP: 11.00Anas × SPP1: 0.940.94TaeniopygiaTaeniopygia × COL1A1: 11.00Taeniopygia × COL1A2: 11.00Taeniopygia × AMELX: 0.10.10Taeniopygia × ENAM: 0.10.10Taeniopygia × AMBN: 0.10.10Taeniopygia × BGLAP: 11.00Taeniopygia × SPP1: 11.00ColumbaColumba × COL1A1: 0.990.99Columba × COL1A2: 0.970.97Columba × AMELX: 00.00Columba × ENAM: 00.00Columba × AMBN: 00.00Columba × BGLAP: 11.00Columba × SPP1: 0.680.68FalcoFalco × COL1A1: 0.910.91Falco × COL1A2: 0.880.88Falco × AMELX: 00.00Falco × ENAM: 00.00Falco × AMBN: 00.00Falco × BGLAP: 0.840.84Falco × SPP1: 0.610.61StruthioStruthio × COL1A1: 0.860.86Struthio × COL1A2: 0.830.83Struthio × AMELX: 0.10.10Struthio × ENAM: 0.10.10Struthio × AMBN: 00.00Struthio × BGLAP: 11.00Struthio × SPP1: 0.570.57AlligatorAlligator × COL1A1: 11.00Alligator × COL1A2: 11.00Alligator × AMELX: 11.00Alligator × ENAM: 0.970.97Alligator × AMBN: 11.00Alligator × BGLAP: 11.00Alligator × SPP1: 11.00CrocodylusCrocodylus × COL1A1: 11.00Crocodylus × COL1A2: 11.00Crocodylus × AMELX: 0.940.94Crocodylus × ENAM: 0.710.71Crocodylus × AMBN: 0.880.88Crocodylus × BGLAP: 11.00Crocodylus × SPP1: 0.810.81GavialisGavialis × COL1A1: 0.930.93Gavialis × COL1A2: 0.890.89Gavialis × AMELX: 0.820.82Gavialis × ENAM: 0.640.64Gavialis × AMBN: 0.770.77Gavialis × BGLAP: 0.860.86Gavialis × SPP1: 0.590.59ChrysemysChrysemys × COL1A1: 11.00Chrysemys × COL1A2: 11.00Chrysemys × AMELX: 0.10.10Chrysemys × ENAM: 0.10.10Chrysemys × AMBN: 0.10.10Chrysemys × BGLAP: 11.00Chrysemys × SPP1: 11.00AnolisAnolis × COL1A1: 11.00Anolis × COL1A2: 11.00Anolis × AMELX: 11.00Anolis × ENAM: 0.950.95Anolis × AMBN: 11.00Anolis × BGLAP: 11.00Anolis × SPP1: 11.00Figure 2. Completeness of extant reference sequences by taxon and proteinFraction of the mature reference chain represented by usable sequence in Library v1.0. Pseudogene remnants (birds, turtle enamel proteins) are shown at 0.1 and are masked in placement. 0 = absent from available assemblies.ProteinCOL1A1COL1A2AMELXENAMAMBNBGLAPSPP1GallusGallus × COL1A1: 11.00Gallus × COL1A2: 11.00Gallus × AMELX: 0.10.10Gallus × ENAM: 0.10.10Gallus × AMBN: 0.10.10Gallus × BGLAP: 11.00Gallus × SPP1: 11.00MeleagrisMeleagris × COL1A1: 11.00Meleagris × COL1A2: 0.980.98Meleagris × AMELX: 0.10.10Meleagris × ENAM: 0.10.10Meleagris × AMBN: 00.00Meleagris × BGLAP: 11.00Meleagris × SPP1: 0.720.72AnasAnas × COL1A1: 11.00Anas × COL1A2: 11.00Anas × AMELX: 00.00Anas × ENAM: 00.00Anas × AMBN: 00.00Anas × BGLAP: 11.00Anas × SPP1: 0.940.94TaeniopygiaTaeniopygia × COL1A1: 11.00Taeniopygia × COL1A2: 11.00Taeniopygia × AMELX: 0.10.10Taeniopygia × ENAM: 0.10.10Taeniopygia × AMBN: 0.10.10Taeniopygia × BGLAP: 11.00Taeniopygia × SPP1: 11.00ColumbaColumba × COL1A1: 0.990.99Columba × COL1A2: 0.970.97Columba × AMELX: 00.00Columba × ENAM: 00.00Columba × AMBN: 00.00Columba × BGLAP: 11.00Columba × SPP1: 0.680.68FalcoFalco × COL1A1: 0.910.91Falco × COL1A2: 0.880.88Falco × AMELX: 00.00Falco × ENAM: 00.00Falco × AMBN: 00.00Falco × BGLAP: 0.840.84Falco × SPP1: 0.610.61StruthioStruthio × COL1A1: 0.860.86Struthio × COL1A2: 0.830.83Struthio × AMELX: 0.10.10Struthio × ENAM: 0.10.10Struthio × AMBN: 00.00Struthio × BGLAP: 11.00Struthio × SPP1: 0.570.57AlligatorAlligator × COL1A1: 11.00Alligator × COL1A2: 11.00Alligator × AMELX: 11.00Alligator × ENAM: 0.970.97Alligator × AMBN: 11.00Alligator × BGLAP: 11.00Alligator × SPP1: 11.00CrocodylusCrocodylus × COL1A1: 11.00Crocodylus × COL1A2: 11.00Crocodylus × AMELX: 0.940.94Crocodylus × ENAM: 0.710.71Crocodylus × AMBN: 0.880.88Crocodylus × BGLAP: 11.00Crocodylus × SPP1: 0.810.81GavialisGavialis × COL1A1: 0.930.93Gavialis × COL1A2: 0.890.89Gavialis × AMELX: 0.820.82Gavialis × ENAM: 0.640.64Gavialis × AMBN: 0.770.77Gavialis × BGLAP: 0.860.86Gavialis × SPP1: 0.590.59ChrysemysChrysemys × COL1A1: 11.00Chrysemys × COL1A2: 11.00Chrysemys × AMELX: 0.10.10Chrysemys × ENAM: 0.10.10Chrysemys × AMBN: 0.10.10Chrysemys × BGLAP: 11.00Chrysemys × SPP1: 11.00AnolisAnolis × COL1A1: 11.00Anolis × COL1A2: 11.00Anolis × AMELX: 11.00Anolis × ENAM: 0.950.95Anolis × AMBN: 11.00Anolis × BGLAP: 11.00Anolis × SPP1: 11.00
Figure 2.Completeness of extant reference sequences by taxon and protein. Fraction of the mature reference chain represented by usable sequence in Library v1.0. Pseudogene remnants (birds, turtle enamel proteins) are shown at 0.1 and are masked in placement. 0 = absent from available assemblies.Source: JM Shenzhen, Peptide Library v1.0 (tag v1.0); public assemblies as of January 2014.
Source data
TaxonCOL1A1COL1A2AMELXENAMAMBNBGLAPSPP1
Gallus110.10.10.111
Meleagris10.980.10.1010.72
Anas1100010.94
Taeniopygia110.10.10.111
Columba0.990.9700010.68
Falco0.910.880000.840.61
Struthio0.860.830.10.1010.57
Alligator1110.97111
Crocodylus110.940.710.8810.81
Gavialis0.930.890.820.640.770.860.59
Chrysemys110.10.10.111
Anolis1110.95111

4.4 Curation

All 214 fragment entries were reviewed under SOP-PL-001 (Table 5). Curators agreed in all fields on first review for 171 entries (79.9%). Agreement on protein assignment alone was higher (Cohen's κ = 0.71) than agreement on acceptance of the computed placement (κ = 0.64). The 43 discordant entries were resolved by discussion in 36 cases and by adjudication in 7.

Table 5. Curation outcomes under SOP-PL-001.
Outcome Entries
Reviewed 214
Concordant on first review (all fields) 171
Discordant on first review 43
— Resolved by curator discussion 36
— Resolved by adjudication 7
Entries carrying a manual override flag 37
— Placement override (computed placement not accepted) 12
— I/L resolution differs from reference residue 11
— Protein assignment differs from search output 9
— PTM annotation differs from search output 5
Override flags with an expiry or review date 0

The twelve placement overrides warrant comment. In each case the computed best placement was within crown Aves or on the crocodilian branch with a posterior below 0.30, and the curators recorded a placement on the avian stem with the justification that the computed placement was weakly supported and inconsistent with the concatenated result. Overridden placements are excluded from the per-fragment statistics in §4.5 and from the concatenated placement; they are retained in the library with their override flags.

Curation required 412 curator-hours in total. Time spent per entry per curator is shown in Fig. 3. The median was 39 minutes; entries in the longest bins were predominantly enamel peptides with partial crocodilian references.

Figure 3. Curator time per library entryRecorded review time per entry per curator under SOP-PL-001, averaged across the two curators, for the 214 fragment entries in Library v1.0. Median 39 minutes. Adjudication time is not included.Entries010203040506070015304560901202400–15: 2215–30: 4830–45: 6145–60: 3960–90: 2790–120: 11120–240: 6Minutes per entryMedian 39 minFigure 3. Curator time per library entryRecorded review time per entry per curator under SOP-PL-001, averaged across the two curators, for the 214 fragment entries in Library v1.0. Median 39 minutes. Adjudication time is not included.Entries0204060800501001502000–15: 2215–30: 4830–45: 6145–60: 3960–90: 2790–120: 11120–240: 6Minutes per entryMedian 39 min
Figure 3.Curator time per library entry. Recorded review time per entry per curator under SOP-PL-001, averaged across the two curators, for the 214 fragment entries in Library v1.0. Median 39 minutes. Adjudication time is not included.Source: JM Shenzhen, curation log CL-2014-01.
Source data
Minutes per entryEntries
0–1522
15–3048
30–4561
45–6039
60–9027
90–12011
120–2406

4.5 Placement on the twelve-taxon reference tree

With the expanded reference set, 147 of 214 fragments are diagnostic (Table 6): 129 of the 131 diagnostic fragments from JM-TR-2013-001 (two were removed by duplicate merging), two that were previously non-diagnostic and now vary among the expanded references, and 16 of the 21 newly admitted fragments. Placement statistics below are computed for the 135 diagnostic fragments without a placement override.

Table 6. Best-supported placement of diagnostic fragments in Library v1.0.
Branch (best-supported) Fragments Of which enamel-derived Median posterior
Stem lineage of Aves 84 4 0.47
Within crown Aves 24 0 0.33
Crocodylia 18 13 0.39
Stem Archosauria 14 8 0.30
Unresolved 7 2 0.22
All diagnostic fragments 147 27 0.41
Of which carrying a placement override 12 3 —

The median per-fragment posterior is 0.41, compared with 0.38 on the seven-taxon tree. The improvement is present at every fragment length but is small (Fig. 4), and the steep dependence of placement confidence on fragment length is unchanged.

Figure 4. Placement posterior by fragment length, seven- versus twelve-taxon referenceMedian per-fragment posterior of the best-supported branch, by fragment length, for diagnostic fragments without a placement override. Seven-taxon tree as in JM-TR-2013-001 (n = 131); twelve-taxon tree, Library v1.0 (n = 135). Band: interquartile range, v1.0.Posterior placement probability0.000.200.400.600.801.00681012141618202224Fragment length (residues)Below 8 residues: high variance7 taxa (JM-TR-2013-001)12 taxa (Library v1.0)Fragment length (residues): 6 7 taxa (JM-TR-2013-001): 0.17 12 taxa (Library v1.0): 0.19 (0.09–0.34)Fragment length (residues): 8 7 taxa (JM-TR-2013-001): 0.26 12 taxa (Library v1.0): 0.29 (0.16–0.43)Fragment length (residues): 10 7 taxa (JM-TR-2013-001): 0.34 12 taxa (Library v1.0): 0.38 (0.24–0.51)Fragment length (residues): 12 7 taxa (JM-TR-2013-001): 0.40 12 taxa (Library v1.0): 0.44 (0.30–0.57)Fragment length (residues): 14 7 taxa (JM-TR-2013-001): 0.46 12 taxa (Library v1.0): 0.50 (0.36–0.62)Fragment length (residues): 16 7 taxa (JM-TR-2013-001): 0.51 12 taxa (Library v1.0): 0.55 (0.41–0.67)Fragment length (residues): 20 7 taxa (JM-TR-2013-001): 0.57 12 taxa (Library v1.0): 0.62 (0.48–0.73)Fragment length (residues): 24 7 taxa (JM-TR-2013-001): 0.62 12 taxa (Library v1.0): 0.67 (0.53–0.78)Figure 4. Placement posterior by fragment length, seven- versus twelve-taxon referenceMedian per-fragment posterior of the best-supported branch, by fragment length, for diagnostic fragments without a placement override. Seven-taxon tree as in JM-TR-2013-001 (n = 131); twelve-taxon tree, Library v1.0 (n = 135). Band: interquartile range, v1.0.Posterior placement probability0.000.200.400.600.801.00101520Fragment length (residues)Below 8 residues:high varianceFragment length (residues): 6 7 taxa (JM-TR-2013-001): 0.17 12 taxa (Library v1.0): 0.19 (0.09–0.34)Fragment length (residues): 8 7 taxa (JM-TR-2013-001): 0.26 12 taxa (Library v1.0): 0.29 (0.16–0.43)Fragment length (residues): 10 7 taxa (JM-TR-2013-001): 0.34 12 taxa (Library v1.0): 0.38 (0.24–0.51)Fragment length (residues): 12 7 taxa (JM-TR-2013-001): 0.40 12 taxa (Library v1.0): 0.44 (0.30–0.57)Fragment length (residues): 14 7 taxa (JM-TR-2013-001): 0.46 12 taxa (Library v1.0): 0.50 (0.36–0.62)Fragment length (residues): 16 7 taxa (JM-TR-2013-001): 0.51 12 taxa (Library v1.0): 0.55 (0.41–0.67)Fragment length (residues): 20 7 taxa (JM-TR-2013-001): 0.57 12 taxa (Library v1.0): 0.62 (0.48–0.73)Fragment length (residues): 24 7 taxa (JM-TR-2013-001): 0.62 12 taxa (Library v1.0): 0.67 (0.53–0.78)
Figure 4.Placement posterior by fragment length, seven- versus twelve-taxon reference. Median per-fragment posterior of the best-supported branch, by fragment length, for diagnostic fragments without a placement override. Seven-taxon tree as in JM-TR-2013-001 (n = 131); twelve-taxon tree, Library v1.0 (n = 135). Band: interquartile range, v1.0.Source: JM Shenzhen, Peptide Library v1.0 (tag v1.0).
Source data
Fragment length (residues)7 taxa (JM-TR-2013-001)12 taxa (Library v1.0)12 taxa (Library v1.0) interval
60.170.190.09–0.34
80.260.290.16–0.43
100.340.380.24–0.51
120.40.440.3–0.57
140.460.50.36–0.62
160.510.550.41–0.67
200.570.620.48–0.73
240.620.670.53–0.78

At the level of individual fragments the change is not uniform. Fig. 5 compares per-fragment posteriors on the two reference trees for a random sample of 24 diagnostic fragments present in both analyses. Most fragments gain modestly; a minority lose support, typically where a new avian reference (Columba, Falco) shares a residue with the fossil fragment and draws placement into crown Aves.

Figure 5. Per-fragment posterior, seven-taxon versus twelve-taxon referencePosterior of the best-supported branch for 24 randomly sampled diagnostic fragments, seven-taxon tree (x) versus twelve-taxon tree (y). Points above the diagonal gained support. Linear fit shown.Posterior, 12-taxon tree0.000.200.400.600.801.000.00.10.20.30.40.50.60.70.80.91.0Posterior, 7-taxon treeDiagnostic fragments linear fit: y = 1.05x + 0.0109, R² = 0.95Diagnostic fragments Posterior, 7-taxon tree: 0.12 Posterior, 12-taxon tree: 0.15Diagnostic fragments Posterior, 7-taxon tree: 0.18 Posterior, 12-taxon tree: 0.17Diagnostic fragments Posterior, 7-taxon tree: 0.21 Posterior, 12-taxon tree: 0.26Diagnostic fragments Posterior, 7-taxon tree: 0.24 Posterior, 12-taxon tree: 0.29Diagnostic fragments Posterior, 7-taxon tree: 0.27 Posterior, 12-taxon tree: 0.22Diagnostic fragments Posterior, 7-taxon tree: 0.29 Posterior, 12-taxon tree: 0.34Diagnostic fragments Posterior, 7-taxon tree: 0.31 Posterior, 12-taxon tree: 0.33Diagnostic fragments Posterior, 7-taxon tree: 0.33 Posterior, 12-taxon tree: 0.39Diagnostic fragments Posterior, 7-taxon tree: 0.35 Posterior, 12-taxon tree: 0.41Diagnostic fragments Posterior, 7-taxon tree: 0.36 Posterior, 12-taxon tree: 0.30Diagnostic fragments Posterior, 7-taxon tree: 0.38 Posterior, 12-taxon tree: 0.42Diagnostic fragments Posterior, 7-taxon tree: 0.4 Posterior, 12-taxon tree: 0.45Diagnostic fragments Posterior, 7-taxon tree: 0.41 Posterior, 12-taxon tree: 0.47Diagnostic fragments Posterior, 7-taxon tree: 0.43 Posterior, 12-taxon tree: 0.44Diagnostic fragments Posterior, 7-taxon tree: 0.45 Posterior, 12-taxon tree: 0.51Diagnostic fragments Posterior, 7-taxon tree: 0.47 Posterior, 12-taxon tree: 0.52Diagnostic fragments Posterior, 7-taxon tree: 0.49 Posterior, 12-taxon tree: 0.41Diagnostic fragments Posterior, 7-taxon tree: 0.52 Posterior, 12-taxon tree: 0.58Diagnostic fragments Posterior, 7-taxon tree: 0.55 Posterior, 12-taxon tree: 0.60Diagnostic fragments Posterior, 7-taxon tree: 0.58 Posterior, 12-taxon tree: 0.63Diagnostic fragments Posterior, 7-taxon tree: 0.61 Posterior, 12-taxon tree: 0.66Diagnostic fragments Posterior, 7-taxon tree: 0.64 Posterior, 12-taxon tree: 0.66Diagnostic fragments Posterior, 7-taxon tree: 0.68 Posterior, 12-taxon tree: 0.73Diagnostic fragments Posterior, 7-taxon tree: 0.72 Posterior, 12-taxon tree: 0.77Figure 5. Per-fragment posterior, seven-taxon versus twelve-taxon referencePosterior of the best-supported branch for 24 randomly sampled diagnostic fragments, seven-taxon tree (x) versus twelve-taxon tree (y). Points above the diagonal gained support. Linear fit shown.Posterior, 12-taxon tree0.000.200.400.600.801.000.00.20.40.60.81.0Posterior, 7-taxon treeDiagnostic fragments linear fit: y = 1.05x + 0.0109, R² = 0.95Diagnostic fragments Posterior, 7-taxon tree: 0.12 Posterior, 12-taxon tree: 0.15Diagnostic fragments Posterior, 7-taxon tree: 0.18 Posterior, 12-taxon tree: 0.17Diagnostic fragments Posterior, 7-taxon tree: 0.21 Posterior, 12-taxon tree: 0.26Diagnostic fragments Posterior, 7-taxon tree: 0.24 Posterior, 12-taxon tree: 0.29Diagnostic fragments Posterior, 7-taxon tree: 0.27 Posterior, 12-taxon tree: 0.22Diagnostic fragments Posterior, 7-taxon tree: 0.29 Posterior, 12-taxon tree: 0.34Diagnostic fragments Posterior, 7-taxon tree: 0.31 Posterior, 12-taxon tree: 0.33Diagnostic fragments Posterior, 7-taxon tree: 0.33 Posterior, 12-taxon tree: 0.39Diagnostic fragments Posterior, 7-taxon tree: 0.35 Posterior, 12-taxon tree: 0.41Diagnostic fragments Posterior, 7-taxon tree: 0.36 Posterior, 12-taxon tree: 0.30Diagnostic fragments Posterior, 7-taxon tree: 0.38 Posterior, 12-taxon tree: 0.42Diagnostic fragments Posterior, 7-taxon tree: 0.4 Posterior, 12-taxon tree: 0.45Diagnostic fragments Posterior, 7-taxon tree: 0.41 Posterior, 12-taxon tree: 0.47Diagnostic fragments Posterior, 7-taxon tree: 0.43 Posterior, 12-taxon tree: 0.44Diagnostic fragments Posterior, 7-taxon tree: 0.45 Posterior, 12-taxon tree: 0.51Diagnostic fragments Posterior, 7-taxon tree: 0.47 Posterior, 12-taxon tree: 0.52Diagnostic fragments Posterior, 7-taxon tree: 0.49 Posterior, 12-taxon tree: 0.41Diagnostic fragments Posterior, 7-taxon tree: 0.52 Posterior, 12-taxon tree: 0.58Diagnostic fragments Posterior, 7-taxon tree: 0.55 Posterior, 12-taxon tree: 0.60Diagnostic fragments Posterior, 7-taxon tree: 0.58 Posterior, 12-taxon tree: 0.63Diagnostic fragments Posterior, 7-taxon tree: 0.61 Posterior, 12-taxon tree: 0.66Diagnostic fragments Posterior, 7-taxon tree: 0.64 Posterior, 12-taxon tree: 0.66Diagnostic fragments Posterior, 7-taxon tree: 0.68 Posterior, 12-taxon tree: 0.73Diagnostic fragments Posterior, 7-taxon tree: 0.72 Posterior, 12-taxon tree: 0.77
Figure 5.Per-fragment posterior, seven-taxon versus twelve-taxon reference. Posterior of the best-supported branch for 24 randomly sampled diagnostic fragments, seven-taxon tree (x) versus twelve-taxon tree (y). Points above the diagonal gained support. Linear fit shown.Source: JM Shenzhen, Peptide Library v1.0 (tag v1.0).
Source data
Posterior, 7-taxon treePosterior, 12-taxon tree
0.120.15
0.180.17
0.210.26
0.240.29
0.270.22
0.290.34
0.310.33
0.330.39
0.350.41
0.360.3
0.380.42
0.40.45
0.410.47
0.430.44
0.450.51
0.470.52
0.490.41
0.520.58
0.550.6
0.580.63
0.610.66
0.640.66
0.680.73
0.720.77

Concatenated, the 135 non-overridden diagnostic fragments place JM-SP-0001 on the stem lineage of Aves with a posterior of 0.93 (ML bootstrap 89%) (Fig. 6). Including the twelve overridden fragments at their curator-assigned placements is not meaningful and was not done.

Figure 6. Library v1.0 reference topology and concatenated placement of JM-SP-0001Fixed twelve-taxon extant reference topology used in Library v1.0, with the concatenated placement of 135 diagnostic fragments from JM-SP-0001 on the stem lineage of Aves (posterior 0.93; ML bootstrap 89%). Chrysemys and Anolis are outgroups.JM-SP-0001p = 0.93JM-SP-0001 — placement posterior 0.93StruthioStruthioGallusGallusMeleagrisMeleagrisAnasAnasColumbaColumbaFalcoFalcoTaeniopygiaTaeniopygiaAlligatorAlligatorCrocodylusCrocodylusGavialisGavialisChrysemysChrysemysAnolisAnolisFigure 6. Library v1.0 reference topology and concatenated placement of JM-SP-0001Fixed twelve-taxon extant reference topology used in Library v1.0, with the concatenated placement of 135 diagnostic fragments from JM-SP-0001 on the stem lineage of Aves (posterior 0.93; ML bootstrap 89%). Chrysemys and Anolis are outgroups.JM-SP-0001p = 0.93JM-SP-0001 — placement posterior 0.93StruthioStruthioGallusGallusMeleagrisMeleagrisAnasAnasColumbaColumbaFalcoFalcoTaeniopygiaTaeniopygiaAlligatorAlligatorCrocodylusCrocodylusGavialisGavialisChrysemysChrysemysAnolisAnolis
Figure 6.Library v1.0 reference topology and concatenated placement of JM-SP-0001. Fixed twelve-taxon extant reference topology used in Library v1.0, with the concatenated placement of 135 diagnostic fragments from JM-SP-0001 on the stem lineage of Aves (posterior 0.93; ML bootstrap 89%). Chrysemys and Anolis are outgroups.Source: JM Shenzhen, Peptide Library v1.0 (tag v1.0).
Source data
TaxonHighlightedPlacement posterior
JM-SP-0001yes0.93
Struthio
Gallus
Meleagris
Anas
Columba
Falco
Taeniopygia
Alligator
Crocodylus
Gavialis
Chrysemys
Anolis

((((JM-SP-0001,(Struthio,(((Gallus,Meleagris),Anas),(Columba,(Falco,Taeniopygia))))),(Alligator,(Crocodylus,Gavialis))),Chrysemys),Anolis);

5. Discussion

Library v1.0 is primarily an infrastructure release. Its scientific content is, by design, almost identical to that of JM-TR-2013-001: one specimen, the same authentication criteria, a modest increase in fragment count from re-analysis of existing data, and a modest improvement in placement from a larger reference set. What has changed is that every fragment now has an identity, a provenance within the data, and a history.

Two results deserve emphasis. First, the reconciliation exercise found that the working files used for JM-TR-2013-001 contained three duplicate peptides and 18 alignment offsets. None of these changes the conclusions of that report; the duplicates inflated the authenticated count by 1.5%, and the offsets affected coverage statistics rather than identifications. They are nevertheless the kind of error that, in an unversioned system, would have propagated silently into any downstream use. Second, the expansion of the reference set improved placement less than the number of added taxa might suggest. The limiting factor for placement is fragment length, not reference density (Fig. 4). Short fragments carry little information regardless of the number of references against which they are compared.

The decision to retain bird and turtle enamel pseudogene remnants, while masking them in placement, may be revisited. They contribute nothing to scoring at v1.0. They are retained because the reconstruction team expects to use conserved positions in their bracketing, and because removal of reference data is not, in our view, a decision to be made at the first release of a library.

The curation procedure is labour-intensive: 412 curator-hours for 214 entries, or roughly two hours per entry including adjudication. This is acceptable for a library of this size. It will not scale linearly, and we expect future specimens to increase the library by an order of magnitude over the coming years. Automation of the concordant cases is the obvious next step, and is out of scope here.

6. Limitations and known issues

  • Single specimen. All 214 fragments derive from JM-SP-0001. All limitations of that specimen described in JM-TR-2013-001 (§6) are inherited by every entry in this release.
  • Manual curation only. No automated consistency checks are applied at build time. Library integrity depends on adherence to SOP-PL-001 by two curators, who are also two of the three authors of this report.
  • Override flags do not expire. Thirty-seven entries carry manual overrides. There is no scheduled review of override flags. Overrides made at v1.0 will remain in force in all subsequent versions unless individually revisited.
  • Placement overrides favour the expected result. All twelve placement overrides move a fragment from a weakly supported unexpected branch to the expected branch. They are excluded from placement statistics for this reason. They are not excluded from the library.
  • Manual bracketing. At v1.0 every fragment is entered into bracketing by hand. This does not scale, and will be replaced by an automated module from v1.1. The automated module's handling of short, modified and I/L-ambiguous fragments has not yet been specified.
  • Override attribution. Override flags are recorded under the shared curation account. The curation log records which pair of curators reviewed an entry, but not which curator proposed an override.
  • Reference quality varies. Falco, Struthio and Gavialis collagen sequences are partial, and several BGLAP and SPP1 sequences derive from draft gene models that have not been independently verified.
  • Enamel references are thin. Four taxa provide usable enamel protein sequence, three of them crocodilians.
  • Public assembly versions. Reference sequences reflect public assemblies as of January 2014. Assembly revisions after that date will not be reflected until a future reference refresh.
  • Infrastructure. The library repository is held on a single server in the Shenzhen laboratory, with nightly backup to a drive in the same room. Build operations are performed under a shared account.

7. Changes in Library v1.0

As published in the Peptide Library Release Notes:

  • First structured release. Consolidates the tooth-derived theropod peptide fragments recovered in 2011–2012, previously maintained as unversioned working files.
  • Initial import of avian and crocodilian reference sequences.
  • Manual curation only.

In detail:

Fragments
214 theropod fragments from specimen JM-SP-0001 (196 from JM-TR-2013-001, less 3 merged duplicates, plus 21 identified by re-search of 2011–2012 raw data). 1,880 unique residues.
References
12 extant taxa (7 avian, 3 crocodilian, 2 outgroup) × 7 proteins. Pseudogene remnants imported and masked.
Non-production set
PL-U, 548 entries, not used in placement or reconstruction.
Curation
SOP-PL-001, two-curator independent review; 37 manual override flags.
Placement
median per-fragment posterior 0.41; concatenated stem-Aves posterior 0.93.

8. Data availability

Library contents are confidential. Library v1.0 is archived as a tagged repository state and remains available to authorized research staff. The 17 original working files are retained read-only with checksums. Raw mass spectrometry data are retained as described in JM-TR-2013-001. No library content is available to third parties.

9. Acknowledgements

We thank the authors and maintainers of the public genome assemblies and protein databases from which the extant reference set was drawn. We thank R. Williams for confirming, to the extent possible, the recorded history of specimen JM-SP-0001.

Funding. This work was funded internally.

Competing interests. All authors are employees of Jurassic Meat and may hold equity.

Ethics. Reviewed by the Ethics & Bioresponsibility Council; 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. 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.
  4. 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.
  5. JM-TR-2013-001. Lin H, Zhou Y, Chen L. Endogenous peptide fragments recovered from an isolated Kem Kem Group theropod tooth: extraction, authentication and preliminary phylogenetic placement. Jurassic Meat technical report, September 2013.

Supplementary Table S1. Fragment entry fields

Table S1. Fields recorded for each fossil-derived fragment entry in Library v1.0.
Field Description
fragment_id Stable identifier, PL-F-NNNNNN; never reused
specimen_id Specimen of origin (JM-SP-0001 for all v1.0 entries)
sequence Peptide sequence; I/L positions marked
il_resolution Residue assigned at each I/L position, and basis
ptm Modifications observed (deamidation, hydroxylation, oxidation, pyro-Glu)
psm_count PSMs at 1% FDR across all campaigns
campaigns Campaigns in which observed (A, B, C)
deamidation_q, deamidation_n Peptide-level deamidated fraction, where measurable
protein Assigned protein (gene symbol)
aln_start, aln_end Coordinates in the per-protein alignment
diagnostic Whether the fragment varies among reference taxa
placement_branch, placement_pp Best-supported branch and posterior
override_flag, override_category, override_note, override_date Manual override record (shared curation account)
curator_1, curator_2, adjudicated_by Curation record
source_files Working files in which the entry appeared before v1.0
first_version Library version in which the entry first appeared