Library v2.8: explicit modelling of diagenetic modification in phylogenetic placement scoring, and deprecation of manual override flags older than 24 months
1 Jurassic Meat Shenzhen Research Center, Shenzhen, China
* Corresponding author: Dr. Hao Lin · regulatory@jurassicmeat.com
Contents
Abstract
Until this release, the Peptide Library scored diagenetically modified residues as if they were sequence. A deamidated asparagine observed as aspartate was treated as an aspartate in the extinct protein. Library v2.8 introduces a diagenesis-aware observation model in which each observed residue is generated from a latent true residue through specimen-specific rates of deamidation (N, Q), oxidation (M, W, H) and in-situ backbone cleavage, with specimen-level diagenetic state informed by Asx racemization where chiral data are available. Applied to the v2.8 library (2,236 theropod-assigned fragments, 12,780 unique residues, 57 extant reference taxa), the model raises median placement posterior from 0.68 to 0.71 and removes the dependence of placement confidence on deamidation load. It reclassifies 1,104 residue calls previously scored as substitutions, including nine of the 31 theropod-diagnostic alignment columns defined in v2.0. The release also inventories manual override records, converted at the v2.0 migration from the v1.x override flags and unflagged manual edits. At the v2.7 freeze there were 4,118 active overrides, 3,342 of them older than 24 months and 1,906 without a recorded rationale. Overrides older than 24 months are deprecated: they emit a build warning and continue to apply pending review. With overrides disabled, median placement posterior is 0.67.
Keywordspalaeoproteomicsdiagenesisdeamidationoxidationracemizationbackbone cleavagephylogenetic placementcuration overridesdata governance
JM-TR-2023-006 · Library v2.8 · 12 September 2023 Dr. Hao Lin*, Dr. Elena Popescu, Dr. Anders Kjeldsen Shenzhen Research Center, Jurassic Meat, Nanshan District, Shenzhen *Corresponding author
Public summary of internal report (full internal version: 46 pages).
1. Summary
- Library v2.8 contains 2,236 theropod-assigned fragments from 21 specimens, covering 12,780 unique residues, against 57 extant reference taxa. Median placement posterior is 0.71.
- Placement scoring now models deamidation, oxidation, backbone cleavage and racemization explicitly. Placement confidence no longer declines with a fragment's deamidation load.
- 1,104 residue calls previously scored as substitutions are now classified as probable diagenetic products, including 9 of 31 theropod-diagnostic alignment columns.
- An inventory found 4,118 active manual override flags in a library of 2,236 fragments. 3,342 were older than 24 months and 1,906 had no recorded rationale.
- Overrides older than 24 months are deprecated. They emit a warning at build time and remain in effect until reviewed. At release, 412 had been reviewed.
2. Introduction
Fossil proteins are not pristine copies of the living protein. Over geological time they undergo hydrolysis of the peptide backbone, deamidation of asparagine and glutamine, oxidation of methionine, tryptophan and histidine, and racemization of amino acids from the L to the D form. These modifications matter in two ways. They are evidence: a fragment showing extensive deamidation and in-situ cleavage is more likely to be old than one that does not, and deamidation has been used as an authenticity criterion in the library since the founding work (JM-TR-2013-001) and in the wider field (Cappellini et al. 2019; Welker et al. 2019). Whether any endogenous protein survives in Mesozoic bone at all remains contested (Schweitzer et al. 2009; Asara et al. 2007; Buckley et al. 2008; Saitta et al. 2019), and the persistence of sequence in closed, mineral-bound systems (Demarchi et al. 2016) is the context in which diagenetic signals are interpreted. They are also noise: a deamidated asparagine has the same elemental composition as aspartate, and a fully deamidated N is indistinguishable by mass from a genuine D.
Through v2.3 the library treated these two roles inconsistently. Modification annotations were stored as a separate layer from v2.0 onward (JM-TR-2019-004, Section 3.3), and were used in contamination screening (JM-TR-2021-005, Section 3.3), but placement scoring used the observed residue. Where an observed D or E sat at a position occupied by N or Q in every reference taxon, the placement engine saw a substitution unique to the fossil and, where that substitution was shared across specimens, a potential synapomorphy. JM-TR-2019-004 and JM-TR-2021-005 both listed diagenetic modelling as a limitation scheduled for a future release. This report describes that release.
The second subject of this report is the manual override flag. Override flags have existed since v1.0, when 37 were applied during manual curation, each with a justification and a date, under a shared curation account and without expiry (JM-TR-2014-002); four more followed in v1.2 (JM-TR-2016-003). At the v2.0 migration the 41 flags were converted to structured override records, together with 1,166 unflagged manual edits in the legacy files, so that manual decisions would be visible (JM-TR-2019-004, CA-6). They have been made visible. Section 4.6 reports what is now visible.
3. Materials and Methods
3.1 Library content
Seven specimens (JM-SP-0051 to JM-SP-0057: three teeth, three long-bone fragments and one ungual) were accessioned between April 2021 and March 2023 under the chain-of-custody schema (JM-TR-2021-005). Extraction, LC-MS/MS and identification followed JM-TR-2019-004 and JM-TR-2021-005 without change, with peptide-level FDR of 1% and MaxQuant-type feature processing (Cox & Mann 2008). The extant reference set was expanded from 41 to 57 taxa, principally by the addition of 14 neognath birds, with one further crocodilian and one further turtle.
3.2 Specimen-level diagenetic profiling
For each specimen we estimated six diagenetic quantities from its accepted PSMs (Fig. 1): fractional deamidation of N and of Q, calculated only at sites where N or Q is conserved in all reference taxa, so that an observed D or E at those sites can be attributed to deamidation with high prior confidence; fractional oxidation of M and of W; and the fraction of peptide termini that are non-tryptic, as a measure of in-situ backbone cleavage. For 14 of 21 specimens, chiral amino acid analysis of a separate subsample provided the D/L ratio of Asx (aspartate plus asparagine) in the acid-insoluble fraction, as an independent measure of thermal and diagenetic history.
These quantities were combined into a single specimen-level diagenetic index θ (0–1) by a latent-variable model fitted across specimens. For the seven specimens without chiral data, θ was estimated from the remaining five quantities.
3.3 Diagenesis-aware observation model
The placement likelihood in v2.3 compared the observed residue at each aligned position with the reference profile directly. In v2.8 the observed residue o is modelled as generated from a latent true residue t through a diagenetic transition matrix D(θ), with the substitution model operating on t:
P(o | tree, placement) = Σ_t P(t | tree, placement) · D_θ(o | t)
D(θ) has non-zero off-diagonal terms for N→D, Q→E, and for oxidised forms of M, W and H. Oxidation does not change the residue identity in the alignment but reduces the confidence weight of the call. Rates are specimen-specific through θ and residue-context-specific through a small set of sequence-context multipliers (for example, the elevated deamidation rate of N followed by G). Explicitly annotated deamidation (where the spectrum localises the mass shift) and inferred deamidation (where a D or E is observed at a conserved N or Q site) are treated identically.
3.4 Cleavage and racemization
In-situ backbone cleavage affects scoring in two ways. First, terminal residues created by diagenetic cleavage, rather than by trypsin, have lower identification confidence, and are down-weighted. Second, the length distribution of diagenetically cleaved fragments is shifted toward short fragments, which place poorly (JM-TR-2019-004, Fig. 2); this is reflected in the per-fragment weight. Racemization is not observed directly at the peptide level. It enters the model only through θ, and through a retention-time term that tolerates the chromatographic shifts associated with isoaspartate formation during deamidation.
3.5 Evaluation
We evaluated the model by ablation (Table 3), by the relationship between deamidation load and placement confidence (Fig. 3), and by the change in per-fragment posterior (Fig. 4). Reconstruction targets were re-inferred under v2.8 scoring and compared with v2.3 scoring.
3.6 Override inventory and deprecation
An override is a structured instruction attached to a library entry that takes precedence over
automated behaviour. Five types exist: FORCE_INCLUDE, FORCE_EXCLUDE, PIN_PLACEMENT, PIN_RESIDUE
and SUPPRESS_CONTAM. Each carries a creator, creation date and an optional rationale. Overrides have
no expiry. We inventoried all overrides active at the v2.7 freeze (31 August 2023) and classified them
by type, age, attribution and rationale. An override is "older than 24 months" if created before
31 August 2021.
4. Results
4.1 Library content
Library v2.8 contains 2,236 theropod-assigned fragments from 21 specimens (Table 1). The 121 keratin-
associated fragments remain flagged (LOWCONF_KAP) and included in default builds at full weight
(JM-TR-2021-005).
| Metric | v2.3 (2021-02) | v2.8 (2023-09) |
|---|---|---|
| Fossil specimens | 14 | 21 |
| Theropod-assigned fragments | 1,610 | 2,236 |
| — collagen type I | 1,219 | 1,677 |
| — enamel proteins | 210 | 303 |
| — other non-collagenous proteins | 94 | 135 |
| — keratin-associated (flagged) | 87 | 121 |
| Unique residues covered | 9,410 | 12,780 |
| Extant reference taxa | 41 | 57 |
| Theropod-diagnostic alignment columns | 31 | 22 (+9 reclassified) |
| Median placement posterior (theropod) | 0.66 | 0.71 |
| Active override flags | 2,690 | 4,024 |
4.2 Diagenetic profiles
Diagenetic profiles vary substantially between specimens (Fig. 1). Q deamidation is lower than N deamidation in every specimen, as expected from their relative rates. The founding specimen JM-SP-0001 is among the most extensively modified. Q deamidation correlates with Asx D/L across the 14 specimens with chiral data (Fig. 2; R² = 0.95), supporting the use of both quantities as measures of a shared diagenetic history rather than of contamination, which would be expected to lower both.
Source data
Specimen N deamidation Q deamidation M oxidation W oxidation Non-tryptic termini Index θ JM-SP-0001 0.91 0.72 0.64 0.48 0.58 0.81 JM-SP-0011 0.89 0.7 0.61 0.46 0.57 0.79 JM-SP-0014 0.86 0.66 0.62 0.44 0.55 0.76 JM-SP-0019 0.83 0.63 0.59 0.41 0.52 0.72 JM-SP-0031 0.82 0.59 0.57 0.39 0.49 0.68 JM-SP-0032 0.79 0.55 0.55 0.37 0.47 0.64 JM-SP-0033 0.69 0.41 0.46 0.29 0.39 0.51 JM-SP-0034 0.71 0.44 0.49 0.31 0.41 0.54 JM-SP-0035 0.77 0.52 0.53 0.36 0.46 0.62 JM-SP-0041 0.84 0.62 0.58 0.4 0.51 0.7 JM-SP-0042 0.66 0.38 0.44 0.27 0.37 0.48 JM-SP-0043 0.58 0.29 0.71 0.55 0.62 0.55 JM-SP-0044 0.61 0.34 0.69 0.52 0.6 0.57 JM-SP-0045 0.72 0.46 0.48 0.32 0.42 0.56 JM-SP-0051 0.87 0.67 0.6 0.43 0.54 0.75 JM-SP-0052 0.74 0.47 0.51 0.33 0.44 0.58 JM-SP-0053 0.68 0.4 0.45 0.28 0.38 0.5 JM-SP-0054 0.92 0.76 0.66 0.5 0.61 0.84 JM-SP-0055 0.73 0.49 0.5 0.34 0.43 0.59 JM-SP-0056 0.81 0.58 0.56 0.38 0.48 0.67 JM-SP-0057 0.63 0.36 0.67 0.51 0.59 0.54
Source data
Label Asx D/L Fractional Q deamidation JM-SP-0043 0.21 0.29 JM-SP-0042 0.26 0.38 JM-SP-0057 0.33 0.36 JM-SP-0033 0.29 0.41 JM-SP-0053 0.35 0.4 JM-SP-0052 0.41 0.47 JM-SP-0034 0.38 0.44 JM-SP-0055 0.44 0.49 JM-SP-0035 0.47 0.52 JM-SP-0032 0.49 0.55 JM-SP-0041 0.55 0.62 JM-SP-0019 0.52 0.63 JM-SP-0014 0.58 0.66 JM-SP-0011 0.61 0.7
The diagenetic model's treatment of each modification class, and the number of residue calls affected in v2.8, is summarised in Table 2.
| Class | Residues | Observable by LC-MS/MS | Treatment in v2.8 | Residue calls affected |
|---|---|---|---|---|
| Deamidation | N→D, Q→E | Partly (mass shift +0.984 Da; indistinguishable from D/E when complete) | Off-diagonal terms in D(θ); conserved-site prior | 1,104 reclassified from substitution |
| Oxidation | M, W, H | Yes (mass shifts) | Reduced confidence weight; identity unchanged | 2,471 down-weighted |
| Backbone cleavage | Any; elevated at Asp–Xaa and Asn | Indirectly (non-tryptic termini) | Terminal residues down-weighted; length term | 3,918 termini down-weighted |
| Racemization | Asx principally | No (requires chiral analysis) | Enters via θ and retention-time tolerance | Not applicable per residue |
4.3 Effect on placement
Under v2.3 scoring, placement confidence declined steeply with the number of deamidated residues in a fragment (Fig. 3). Under v2.8 scoring, the decline is largely removed. The effect is concentrated in the 486 fragments with two or more deamidated residues, which are also those with the strongest evidence of age.
Source data
Deamidated residues per fragment v2.3 scoring v2.8 scoring v2.8 scoring interval 0 0.7 0.72 0.55–0.85 1 0.66 0.71 0.54–0.84 2 0.58 0.69 0.51–0.83 3 0.49 0.66 0.47–0.81 4 0.41 0.62 0.41–0.79 ≥5 0.33 0.57 0.34–0.77
Per-fragment posterior increased for 1,561 fragments and decreased for 675 (30.2%) (Fig. 4). Decreases are concentrated in fragments whose placement had depended on a D or E residue at a reference-conserved N or Q site, that is, on a residue the model now attributes to deamidation rather than to inheritance.
Source data
Change in posterior placement probability Fragments −0.3–−0.2 21 −0.2–−0.1 88 −0.1–−0.05 164 −0.05–0 402 0–0.05 917 0.05–0.1 341 0.1–0.2 212 0.2–0.3 71 0.3–0.4 20
The ablation in Table 3 separates the contributions of the reference-set expansion, the diagenetic model and the override flags to the headline median. The reported figure for v2.8 (0.71) is computed with all active overrides applied, consistent with every previous release. Re-scored under the v2.8 model, the v2.3 entry set has a median of 0.68 (run A′); the v2.3 figure as released (0.66) is not restated.
| Run | Library | Reference taxa | Scoring | Overrides applied | Median posterior |
|---|---|---|---|---|---|
| A | v2.3 | 41 | v2.3 | All | 0.66 |
| A′ | v2.3 | 41 | v2.8 | All | 0.68 |
| B | v2.8 | 57 | v2.3 | All | 0.68 |
| C | v2.8 | 57 | v2.8 | All | 0.71 (reported) |
| D | v2.8 | 57 | v2.8 | Only those ≤24 months old | 0.69 |
| E | v2.8 | 57 | v2.8 | None | 0.67 |
| F | v2.8 | 57 | v2.3 | None | 0.64 |
The consensus placement under v2.8 scoring remains on the avian stem (Fig. 5).
Source data
Taxon Highlighted Placement posterior JM theropod yes 0.71 Struthio Rhea Dromaius Tinamus Gallus Meleagris Anas Columba Falco Taeniopygia Alligator Caiman Crocodylus Gavialis Chrysemys (((JM_theropod,((Struthio,(Rhea,(Dromaius,Tinamus))),(((Gallus,Meleagris),Anas),(Columba,(Falco,Taeniopygia))))),((Alligator,Caiman),(Crocodylus,Gavialis))),Chrysemys);
4.4 Theropod-diagnostic columns
Library v2.0 defined 31 alignment columns at which theropod-assigned fragments carry a residue absent
from all extant reference taxa (JM-TR-2019-004, Table 1). These columns have been cited internally as
the most direct sequence-level evidence of a non-avian theropod source. Under v2.8 scoring, nine of the
31 are reclassified as probable deamidation products: in each, the theropod residue is D or E and every
one of the 57 reference taxa carries N or Q (Table 4). The remaining 22 columns are retained as
diagnostic. Seven of the nine reclassified columns carry PIN_RESIDUE overrides, created between
2019 and 2021, which fix the theropod residue as D or E. Overrides take precedence over the model
(Section 3.6), and in default builds these seven columns therefore continue to be scored as diagnostic.
| Column | Locus | Reference residue (57/57 taxa) | Theropod residue | Fragments | Deamidation posterior | PIN_RESIDUE override |
|---|---|---|---|---|---|---|
| C1A1-0412 | COL1A1 | N | D | 38 | 0.97 | Yes (2019) |
| C1A1-0587 | COL1A1 | Q | E | 22 | 0.91 | Yes (2019) |
| C1A1-0903 | COL1A1 | N | D | 41 | 0.98 | Yes (2019) |
| C1A2-0219 | COL1A2 | N | D | 17 | 0.94 | Yes (2020) |
| C1A2-0656 | COL1A2 | Q | E | 13 | 0.86 | No |
| C1A2-0871 | COL1A2 | N | D | 29 | 0.96 | Yes (2020) |
| AMLX-0044 | AMELX | Q | E | 11 | 0.88 | Yes (2021) |
| ENAM-0130 | ENAM | N | D | 6 | 0.83 | No |
| C1A1-1024 | COL1A1 | N | D | 19 | 0.95 | Yes (2021) |
4.5 Effect on reconstruction targets
Re-inference of all 19 current reconstruction targets under v2.8 scoring changed inferred residues at 12 positions in five targets. Three changes fall in targets incorporated in the production cell line DN-2019-04 (JM-TR-2019-004); all three are at sites with posterior below 0.6 under both scorings and lie outside the regions characterised for that line in 2021. Impact assessment IA-2023-03 concluded that there was no impact on released material. The assessment was reviewed by the Ethics & Bioresponsibility Council on 22 August 2023: reviewed; no concerns raised.
4.6 Override inventory
At the v2.7 freeze there were 4,118 active override flags in a library of 2,236 fragments (Table 5).
Overrides are not evenly distributed: 61% apply to 14% of fragments, principally those from JM-SP-0001
and the ungual specimens. Of the 4,118, 3,342 (81.2%) were older than 24 months, 1,906 (46.3%) had
no recorded rationale, and 1,207 (29.3%) were legacy records converted at the v2.0 migration: the 41 flags carried from v1.x and
1,166 records inferred from unflagged manual edits in the legacy files, all attributed to the migration
service account jm-curate because individual attribution was not recorded at the time (JM-TR-2019-004,
Section 6).
| Type | Effect | Active | Older than 24 months | No recorded rationale |
|---|---|---|---|---|
FORCE_INCLUDE |
Entry included regardless of filters and exclusion rules | 1,184 | 961 | 502 |
FORCE_EXCLUDE |
Entry excluded regardless of scoring | 391 | 318 | 140 |
PIN_PLACEMENT |
Placement fixed to a specified branch | 742 | 655 | 377 |
PIN_RESIDUE |
Residue at a column fixed to a specified value | 1,513 | 1,214 | 742 |
SUPPRESS_CONTAM |
Contamination-database match disregarded | 288 | 194 | 145 |
| Total | 4,118 | 3,342 | 1,906 |
Further findings from the inventory:
- 388
PIN_RESIDUEoverrides fix a D or E at a position the v2.8 model classifies as probable deamidation (posterior > 0.8). Seven of these are the diagnostic columns in Table 4. - 41
SUPPRESS_CONTAMoverrides apply to keratin-associated fragments, including the four described in JM-TR-2021-005, Section 4.2. - 63 fragments carry both a
FORCE_INCLUDEand aFORCE_EXCLUDEoverride. These are resolved by the build system's precedence rule, under which the more recent override applies. This rule is documented here for the first time. - 17 overrides reference fragment identifiers that do not exist in the library, as a result of the renumbering at the v2.0 migration. They have no effect and have been retained for audit.
- In v1.x, fragments with placement overrides were excluded from per-fragment placement statistics (JM-TR-2014-002). Since v2.0 they have been included at their pinned placement. This change was made at migration to preserve build equivalence and was not separately documented.
PIN_PLACEMENToverrides contribute 0.02 of the headline median posterior (Table 3, runs C and D) and all override types together contribute 0.04 (runs C and E).
The age profile of overrides is shown in Fig. 6. The largest single year of creation is 2021, which coincides with the curation of the keratin-associated peptides for v2.3.
4.7 Deprecation and review
Overrides older than 24 months are deprecated in v2.8. A deprecated override emits warning W-OVR-24M
at build time and continues to apply. It is removed only when reviewed and retired; it ceases to be
deprecated when reviewed and renewed with a recorded rationale, which is now mandatory for renewal.
Review began on 3 July 2023. Its status at release is shown in Table 6.
| Status | Overrides |
|---|---|
| Older than 24 months at v2.7 freeze | 3,342 |
| Reviewed and renewed (rationale recorded) | 318 |
| Reviewed and retired | 94 |
| Not yet reviewed (deprecated; in effect) | 2,930 |
| Active overrides at v2.8 release | 4,024 |
W-OVR-24M warnings emitted by a default v2.8 build |
2,930 |
At the current review rate, which is limited by the need to re-examine raw spectra for many
PIN_RESIDUE and SUPPRESS_CONTAM overrides, review of the remaining deprecated overrides is expected
to take approximately 30 months. All build warnings are written to the build log, which is retained
with each release (JM-TR-2019-004, CA-4).
5. Discussion
The diagenesis-aware model does what it was designed to do. Placement confidence is no longer penalised for the modification that most strongly indicates age, and the headline median rises from 0.68 to 0.71 on the same library and reference set. The direction of this result is reassuring: the most heavily modified fragments, those least likely to be modern contaminants, were the ones the previous scoring placed least well, and are now placed comparably to unmodified fragments.
The reclassification of nine diagnostic columns requires a more careful statement. Each of those
columns shows a D or E in the fossil where all 57 living relatives carry N or Q. That is the pattern
expected from a genuine substitution on the theropod lineage, and it is also exactly the pattern
expected from deamidation. The previous scoring could not distinguish the two and did not try. The v2.8
model assigns each a deamidation posterior between 0.83 and 0.98. Placement on the avian stem is the
expectation under the extant phylogenetic bracket (Witmer 1995). We do not regard this as evidence
against a theropod source, which rests on the 22 remaining columns and on the joint placement of
2,236 fragments, but these nine columns should no longer be cited as diagnostic. That seven of them
continue to be scored as diagnostic in default builds, through PIN_RESIDUE overrides created before
the diagenetic model existed, illustrates the governance problem addressed in the second half of this
release.
Overrides were introduced to make manual decisions visible and attributable. The inventory shows that they are visible. It also shows that they accumulate: the library now carries more overrides than fragments, almost half without a rationale, and the oldest and most numerous were inherited from a period in which curation decisions were not attributed to individuals. Overrides are not in themselves errors. Many of those reviewed so far were renewed, and a curation decision taken with the spectrum in front of the curator may be better than any automated rule. However, an override has no expiry, takes precedence over every subsequent improvement to the scoring, and, as runs C to E in Table 3 show, contributes measurably to the confidence figure we report. A 24-month deprecation threshold is a first step. We chose deprecation with warnings rather than automatic expiry because automatic expiry would change 3,342 entries at once, and the v2.0 post-mortem gave us a durable aversion to large unexamined changes in build behaviour.
6. Limitations and known issues
- Overrides take precedence over the diagenetic model. 388
PIN_RESIDUEoverrides fix a residue the model attributes to deamidation, including seven of nine reclassified diagnostic columns. - 2,930 deprecated overrides remain in effect pending review. A default build emits 2,930 warnings.
- The reported median posterior (0.71) includes the effect of overrides. Without overrides it is 0.67.
- 63 fragments carry contradictory include/exclude overrides, resolved by recency.
- 17 overrides reference fragments that do not exist.
- Racemization data are available for 14 of 21 specimens; θ is imputed for the remainder. JM-SP-0001 was not subsampled for chiral analysis: its remaining crown material is reserved under the foundational-specimen provision (JM-TR-2016-003).
- The model assumes all modification is post-mortem. Deamidation accumulating in long-lived proteins during the life of the animal is not distinguished.
- Deamidation rates are context-dependent at a finer scale than the model's context multipliers capture, including dependence on local structure that is not available for fossil material.
- Keratin-associated fragments (121) remain flagged and included at full weight.
- Fragments shorter than eight residues remain poorly placed under both scorings.
- Clade-targeted build filters remain user-specified and are not validated against the target clade. A mis-set filter is not detected at build time.
7. Changes in Library v2.8
- Improved handling of diagenetically altered residues.
- Deprecated manual override flags older than 24 months.
Also listed in the release notes for v2.8:
- 626 theropod-assigned fossil-derived entries added (1,610 → 2,236); sixteen extant reference taxa added (41 → 57).
- Unaligned legacy peptide files (deprecated in v2.0) removed. All legacy content had been converted.
- Fixed: the contamination exclusion list was not applied to identifications supported by de novo sequencing alone. Affected entries were re-screened; the reclassified entries are listed in the v2.8 validation report. No affected entry contributed to a production target sequence.
- Fixed: the build manifest reported reference taxa counts that included two duplicate species entered under a former and a current scientific name. Corrected before release; the released count (57) excludes the duplicates.
- Known issue: racemization estimates are unavailable for seven of 21 specimens, including JM-SP-0001 (KI-0519; Section 6).
The diagenetic model is described in Sections 3.2–3.4 and 4.2–4.5; override deprecation in Sections 3.6 and 4.6–4.7.
8. Data availability
Library contents are confidential. Library access is restricted to authorized research staff and is not available to third parties. The override register and build logs are available to auditors and regulators on request under confidentiality. Aggregate statistics are derived from build manifest v2.8, diagenetic profiling run DG-2023-02 and the override register snapshot of 31 August 2023. Questions about this report may be directed to regulatory@jurassicmeat.com.
9. Acknowledgements
We thank the mass spectrometry suite staff for the re-examination of archived spectra during override review, Marcus Oyelaran for coordinating subsampling for chiral analysis under the custody procedure, and Dr. Rahul Venkataraman for re-inference of reconstruction targets. We thank the curators, past and present, whose decisions are inventoried in Section 4.6. The override review and IA-2023-03 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
- Asara JM, et al. (2007). Protein sequences from mastodon and Tyrannosaurus rex revealed by mass spectrometry. Science 316:280–285.
- Buckley M, et al. (2008). Comment on "Protein sequences from mastodon and Tyrannosaurus rex revealed by mass spectrometry." Science 319:33.
- Cappellini E, et al. (2019). Early Pleistocene enamel proteome from Dmanisi resolves Stephanorhinus phylogeny. Nature 574:103–107.
- 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.
- Demarchi B, et al. (2016). Protein sequences bound to mineral surfaces persist into deep time. eLife 5:e17092.
- Saitta ET, et al. (2019). Cretaceous dinosaur bone contains recent organic material and provides an environment conducive to microbial communities. eLife 8:e46205.
- Schweitzer MH, et al. (2009). Biomolecular characterization and protein sequences of the Campanian hadrosaur B. canadensis. Science 324:626–631.
- Welker F, et al. (2019). Enamel proteome shows that Gigantopithecus was an early diverging pongine. Nature 576:262–265.
- 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.
- JM-TR-2013-001. Endogenous peptide fragments recovered from a Kem Kem Group theropod tooth. Jurassic Meat internal technical report.
- JM-TR-2019-004. Library v2.0: a versioned multiple-sequence-alignment architecture for fragmentary ancient peptides. Jurassic Meat internal technical report.
- JM-TR-2021-005. Library v2.3: keratin-associated peptides and a chain-of-custody metadata schema. Jurassic Meat internal technical report.