Human serum peptidase calibration audit¶
This audit addresses #543. It packages 12 study records and 27 reported hormone-stability measurements, with sources, assay conditions, perturbations, reuse terms and identifiability. The research result is a bounded evidence inventory and minimum calibration experiment. The audited inputs do not justify a new fit of transferable enzyme hazards for human long vaccine peptides and their fragments.
What was found¶
| Evidence | Available input | What it identifies |
|---|---|---|
| Abid 2009, human serum oxidized NPY | Km/Vmax, inhibitors, products | Specific apparent DPP4/aminopeptidase-P/kallikrein reference rates; chemistry, pH 8 and manganese remain part of the assay |
| Kuoppala 2000, human plasma bradykinin | Concentration-dependent product attribution | ACE and CPN-like pathway dominance bounds, without transferable vaccine hazards |
| Yi 2015, matched human collection tubes | Table 2: 27 half-life records; MS/antibody time-course figures; six TIF supplements | Assay-specific peptide-signal stability and cocktail protection; individual enzyme contributions remain unresolved |
| Zhen 2016, human plasma FGF21 | Selective FAP inhibitor, depletion and rescue; LC-MRM products and plotted trajectories | Causal support for FAP's C-terminal cleavage of FGF21; a 181-aa protein and plotted endpoints do not calibrate vaccine peptide hazards |
| Bainbridge 2017, FAP probes | Human-serum depletion, animal knockout and purified-enzyme comparisons | FAP reporter specificity; ordinary GP probes are also cleaved by PREP. Selective D-Ala probes have dyes and noncanonical chemistry |
| Toräng 2016, human PYY | In-vitro products and in-vivo sitagliptin arm | Fragment and assay-endpoint differences; sampling inhibitors are quench reagents in vitro, and contaminated infusate prevents a reliable in-vivo conversion-rate estimate |
| Nyimanu 2019, human apelin | In-vivo identified metabolites | Products compatible with proposed MME activity; no matched MME perturbation to isolate its hazard |
| Dufresne 2017, broad human plasma peptidomics | Supplemental peptide/protein matrix, precursor intensities and identification frequency | Preanalytical proteolysis; ice/inhibitor versus warm samples confound temperature and inhibition, and products mix generation with loss |
| Yi 2007, human serum/plasma | Broad inhibitor collection devices and endogenous fragments | Sequential processing and anticoagulant effects; numerical trajectories/licensing were not established in this audit |
| FTMS 2007 | Fractionated human peptide pool plus exogenous DPP4/APP2 | Substrate discovery and a purified DPP4 kinetic validation; intact-matrix kinetics remain unmeasured |
| Maffioli 2020 | Pig-plasma 228-peptide library and numerical products | Broad boundary evidence, with no selective named-enzyme calibration or validated human transfer |
| Böttger 2017 | Mouse blood/plasma/serum comparison | Matrix/coagulation can alter mechanism and stability ranking; no human vaccine enzyme calibration |
The machine-readable inventory records inspected downloads and hashes where available. API failures or unverified supplements stay explicit. A public supplement's existence is distinguished from successful local inspection. Searches covered human matrix degradation/inhibitors, DPP4/ACE/FAP/MME, matched peptidomics and MSP-MS libraries; this audit is not proof that no other dataset exists. Original figures and article prose are not redistributed.
Reproduce reported reference clocks¶
from mhctools import serum_calibration_evidence, serum_assay_parent_reference
records = serum_calibration_evidence()["measurements"]
reference = serum_assay_parent_reference("yi2015:t2:05", [0, 4, 8])
This record is GLP-1 G36A signal in human serum at room temperature, reported as 4 ± 0.2 hours. Under the source's first-order assumption its reference signal fractions are 1, 0.5 and 0.25. This is a reproduction of a reported parent clock, without fitting raw trajectories or assigning enzyme hazards. Amidation, matrix, assay and temperature remain explicit. Antibody-only and combined MS/antibody rows retain their observable/concentration ambiguity.
>96 h, <24 h and donor ranges remain bounds with no midpoint. A replicate
SD is separate from those bounds and does not become a confidence interval.
The 96-hour horizon is the study's maximum; individual row horizons were not
reported in the table and stay unknown.
No bond allocation, target-survival trajectory or vaccine sequence lookup is
supplied by this reference. Original NPY enzyme-specific reference calculations
remain in serum contributions.
Source inconsistencies are recorded: Table 2, methods and body report different room-temperature ranges; the glucagon EDTA range differs between table and body; OXM-K33 in methods differs from generic OXM table labels. Table 2 values are retained with these notes. Repeated assay/donor records are not independent training examples.
python scripts/serum_calibration_review.py --output /tmp/serum-calibration \
--source-xml /path/to/PMC4519045.xml
This audits every Table 2 row, including rowspans, censoring and uncertainty, and exports source-linked JSON/CSV with checksums.
Minimum vaccine-specific calibration experiment¶
Use the actual peptide chemistry, selected epitope and observed target-bearing fragments. Compare vehicle with each selective inhibitor at the same temperature, substrate concentration, matrix fraction and donor preparation. Record serum/plasma processing, anticoagulant, pH, enzyme activation, inhibitor identity/dose/selectivity and sample IDs. Include orthogonal depletion/rescue and combination arms when needed to distinguish shared products and serial pathways.
Measure intact parent and identified target-bearing products over sufficiently dense early and late time points, using calibrated LC-MS, recovery controls, stable-isotope standards and LOD/LOQ. Multiple donors and concentrations are needed to assess donor variation and concentration dependence. Publish the numerical trajectories and source product identities.
Fit only identifiable competing/serial substrate-scoped rates. Hold out donors or sequences; separate parameter uncertainty from missing mechanisms. Report changes in exact-target availability on enzyme removal without renormalizing remaining hazards. MHC presentation and APC uptake need separate measurements.
For the shortest interpretation, use the three-route summary.