Peer-review response · validation protocol · v0.1

Tiramisù Spoilage Simulator — Revised Conclusions & Test Path

Response to an independent reviewer: porous biscuit air, temperature practicality, cocoa bioburden, natamycin legality, nutrient-factor calibration, packaging bioburden, and the meaning of “reported” shelf-life days.

§ 01Executive revision to the scientific conclusions

Conclusion that should change most

Natamycin must be demoted from “recommended lever” to “technical positive control / jurisdiction-dependent option”. In EU-style regulatory framing, its use on tiramisù surfaces is a likely showstopper unless specifically authorised. The study should instead prioritise legal, taste-neutral levers: cocoa bioburden, packaging hygiene, fill-air control, barrier packaging, and robust cold-chain verification.

Conclusion that needs new experiments

Cocoa bioburden is under-isolated in the current scenario set. It appears in mixed scenarios where air, cups, temperature, barrier lid and natamycin also change. Add independent cocoa-ladder tests with sterile/gamma/VHP cups and clean filling, so cocoa quality can be quantified as its own lever.

Bottom line after peer review The model remains useful, but the manuscript should pivot from “best alcohol-free path = 2 °C + natamycin + packaging” toward “best validated path = independent control of cocoa, packaging surfaces and airborne filling contamination; temperature and natamycin are high-impact but constrained levers.”

§ 02Point-by-point response to reviewer

Reviewer issueDoes current work address it?What it adds to conclusionsRequired revision
Air flux from porous biscuits Partly. The simulator includes V_biscuit_pore and f_trapped_air as an initial trapped-air reservoir, but not time-dependent gas flux from biscuit pores during storage. Relevant mainly for N₂/MAP/barrier scenarios. It may explain oxygen rebound after sealing and reduce the apparent benefit of N₂ flush. Add a dynamic pore-gas exchange term or explicitly state “initial dilution only”. Validate with O₂ sensors in cups with normal vs vacuum-degassed biscuits.
Lowering temperature is difficult Current report treats 2 °C as a practical lever. Revise: 2 °C is a high-impact sensitivity / premium controlled-chain option, not a generally flexible industrial setting. Oscillations can cause local freezing or texture damage. Keep P06 but relabel as “hypothetical or controlled cold-chain scenario”; include stable 4 °C, stable 2 °C and oscillating 0–4 °C validation arms.
Cocoa bioburden not independently tested True. Cocoa changes are mixed with other changes in P05/P09/P10. Cocoa deserves higher priority than natamycin because it is a legal ingredient-specification lever. Add cocoa ladder: sterile/irradiated cocoa, steam-treated cocoa, standard cocoa, intentionally high-bioburden cocoa.
Nutrient factor now very relevant Yes. The simulator uses nutrient factors of 0.2 for cocoa, 0.3 for lid underside, 0.5 for cup rim, and 1.0 for bulk/cream-contact surfaces. The model’s compartment ranking depends strongly on nutrient assumptions. This is acceptable only if treated as a calibration parameter. Validate with surface-coupon tests: clean lid, rim with splatter, cocoa layer, cream smear; fit nutrient factors from radial growth rates.
Airborne control vs cup/lid bioburden Current language mixes them. The simulator has separate inputs for airborne CFU/cup and package CFU/cm², but the report’s “fix the plant” wording is ambiguous. Separate levers: fill-air deposition, cup internal surface, lid product-facing surface, and ingredient/cocoa bioburden. Rename “airborne contamination control” to “fill-air deposition control” and add “packaging-surface bioburden control” as a distinct intervention.
“Reported” days look too precise The report defines reported days as predicted × 0.7 safety factor, not measured shelf life. The label is misleading. It can be mistaken for real market data matching the scenario. Rename to “conservative model output” or “screening recommendation after 0.7 factor”. Round to whole days or half-days.

§ 03Revised hierarchy of practical levers

RankLeverPractical statusWhy it moves up or downValidation priority
1Cup/lid surface bioburden controllegal & practicalDirectly affects rim/lid limiters. Can be tested with VHP, UV, gamma-irradiated packaging, sterile bags, and swabs.critical
2Fill-air deposition controllegal & practicalSeparate from packaging. Cleanroom/laminar flow can reduce spores deposited during open-cup dwell time.critical
3Cocoa bioburden/specificationlegal & practicalMore publishable and deployable than natamycin; must be isolated from packaging and air effects.critical
4Nutrient availability on lid/rim/cocoamodel-sensitiveNot a process lever alone, but determines whether rim/lid/cocoa predictions are credible.critical
5Barrier lid + N₂ flushlegal & practicalUseful but likely limited if biscuit pores and residual O₂ dominate. Needs pore-air validation.high
62 °C cold chainhigh impact, constrainedBiologically powerful, but often difficult commercially; freezing/texture risk under normal oscillations.high
7Natamycinpossible showstopperTechnically strong antifungal, but food-additive authorisation may not cover tiramisù surfaces.optional

§ 04Proposed validation path — feasible lab programme

PHASE 0Instrument the product

Measure aw, pH, O₂/CO₂, temperature, texture/freezing indicators, and initial bioburden.

PHASE 1Calibrate components

Validate cocoa, packaging, fill-air, biscuit-pore and nutrient-factor submodels independently.

PHASE 2Run factorial cup tests

Assemble real cups under cleanroom/sterile-equipment conditions and follow visible mold/time-to-failure.

PHASE 3Fit and challenge

Calibrate parameters on one data subset, then blind-predict another subset.

Phase 0 — baseline measurements before microbiology

MeasurementMethodReasonDecision threshold
aw by compartmentaw meter on cocoa, top cream, biscuit, bottom cream at day 0 and after equilibrationNeeded because cocoa aw is modelled via thin-layer coupling, not simply user input.If cocoa aw differs by >0.03 from model, refit cocoa L0_mm.
pH by compartmentMicroelectrode or homogenised compartment pHWeak-acid preservatives and pathogen boundaries depend on pH.If pH drifts >0.2 units, transport/buffering assumptions need revision.
Headspace O₂/CO₂Non-destructive optical spot sensors or septum samplingValidates OTR, trapped pore air and N₂ flush persistence.If O₂ rebounds faster than lid OTR predicts, biscuit-pore flux is relevant.
Temperature distributionData loggers inside dummy cups, lid surface, carton centreNeeded before treating 2 °C as realistic.Any local sub-zero episodes require texture/freezing assessment.
Texture/freezingVisual ice check, syneresis, cream rheology, sensory screenLow-temperature shelf-life gain is useless if quality fails.Reject 2 °C scenario if oscillations damage cream or biscuit texture.

Phase 1 — independent calibration experiments

Test blockExperimental designControlsPrimary endpointModel parameter tested
A. Cocoa bioburden ladder Prepare identical sterile-base cups using four cocoa levels: sterile/irradiated, steam-treated, standard supplier, and high-bioburden inoculated cocoa. Gamma or VHP-sterilised cups/lids; cleanroom filling; no natamycin; same 4 °C storage. Time to visible cocoa mold and initial mold spores/g. Cocoa source term, cocoa nutrient factor, cocoa aw coupling.
B. Cup/lid surface bioburden Use the same sterile formulation and cocoa, but compare gamma-irradiated packaging, VHP-sterilised packaging, supplier-clean packaging, and deliberately contaminated packaging coupons. Cleanroom fill and low-bioburden cocoa fixed. Rim/lid mold incidence and location. Package CFU/cm² inputs and rim/lid limiters.
C. Fill-air deposition Same ingredients and packaging, filled under ISO/laminar flow, clean room, and controlled open-air exposure. Use settle plates or active air sampling during filling. Gamma/VHP packaging fixed; cocoa fixed. Airborne CFU/cup and resulting surface mold time. Airborne deposition partitioning across cocoa/rim/lid.
D. Porous biscuit air N₂-flushed cups with normal soaked biscuits, vacuum-degassed soaked biscuits, fully saturated biscuits, and dry/high-pore biscuits. Barrier lid fixed; sterile filling; non-destructive O₂ sensors. O₂ rebound curve over 0–7 days and mold time under N₂. V_biscuit_pore, f_trapped_air, and any new pore-exchange constant.
E. Nutrient-factor surface coupons Inoculate identical spores onto lid film, cup rim material, cocoa layer, cream smear and cocoa-over-cream coupons under controlled humidity. Same strain/spore dose, temperature and gas atmosphere. Radial growth rate, germination time, visible colony time. nutrient_factor for cocoa/rim/lid.
F. Cold-chain realism Store identical cups at stable 4 °C, stable 2 °C, oscillating 0–4 °C and abuse 8 °C. Same batch, packaging, cocoa and air conditions. Mold time plus texture/freezing/syneresis. Temperature gamma; practical feasibility of 2 °C.
G. Natamycin ring-fenced control Only if allowed by the target jurisdiction or run as a non-commercial technical control. Compare 0 vs 5 vs 10 ppm surface exposure on coupons/cups. Same cocoa, packaging and air conditions. Fungal inhibition without bulk sensory change. Natamycin gamma; not a default recommendation.
Feasible minimum If resources are limited, prioritise blocks A–E. They directly address the reviewer’s criticisms and validate the strongest scientific claims. Block F is important for practical recommendations; block G is optional unless the regulatory path is confirmed.

§ 05Suggested cup-level validation matrix

This matrix keeps the experiment feasible by using a staged fractional design rather than every possible combination.

IDPurposeCocoaPackagingFill environmentGas/lidTemperatureExpected learning
V1Clean baselineSteam-treatedVHP/gamma sterileCleanroomAir / PP4 °CLowest practical no-preservative baseline.
V2Cocoa sterileSterile/irradiatedVHP/gamma sterileCleanroomAir / PP4 °CResidual failures are not cocoa-driven.
V3Cocoa standardStandard supplierVHP/gamma sterileCleanroomAir / PP4 °CIndependent cocoa effect.
V4Cocoa highHigh-bioburden/inoculatedVHP/gamma sterileCleanroomAir / PP4 °CDose-response and model slope.
V5Cup/lid supplier effectSteam-treatedSupplier-cleanCleanroomAir / PP4 °CPackaging bioburden contribution.
V6Dirty package positiveSteam-treatedKnown contaminatedCleanroomAir / PP4 °CRim/lid failure sensitivity.
V7Airborne effectSteam-treatedVHP/gamma sterileStandard room/open fillAir / PP4 °CFill-air deposition contribution.
V8N₂ with normal biscuitSteam-treatedVHP/gamma sterileCleanroomN₂ / barrier4 °CActual benefit of inert headspace.
V9N₂ with degassed biscuitSteam-treatedVHP/gamma sterileCleanroomN₂ / barrier4 °CQuantifies porous-biscuit O₂ contribution.
V10Low-T idealSteam-treatedVHP/gamma sterileCleanroomAir / PP2 °C stableBiological temperature ceiling.
V11Low-T realisticSteam-treatedVHP/gamma sterileCleanroomAir / PP0–4 °C oscillatingFreezing/quality risk and practical limit.
V12Technical natamycin controlSteam-treatedVHP/gamma sterileCleanroomAir / PP4 °COnly if legally/ethically permitted; calibrates antifungal upper bound.

Replication and sampling

§ 06Model changes before re-running scenarios

ChangeWhyImplementationEffect on paper
Rename “reported days”Avoid implying real market or experimental data.Use “conservative screening days = predicted × 0.7”, rounded to whole days.Major credibility gain.
Split contamination leversAirborne ≠ cup/lid surface bioburden.Separate report sections: fill-air deposition, cup interior, lid product-facing, cocoa, ingredients.Clearer practical recommendations.
Add biscuit pore flux optionCurrent model only mixes trapped air at t=0.Add optional k_pore exchange between biscuit pore reservoir and headspace, fitted from O₂ curves.Better MAP/N₂ predictions.
Make nutrient factors calibration parametersThey strongly influence rim/lid/cocoa ranking.Report sensitivity tornado; fit values from coupon radial growth.Prevents overclaiming.
Add cocoa-only scenariosIndependent cocoa effect missing.P13–P16 cocoa ladder at fixed clean packaging/air.Moves study toward deployable conclusions.
Demote natamycin in strategyPotential additive-law showstopper.Move to appendix or “technical positive control”.Avoids regulatory criticism.
Relabel 2 °CCold chain usually not flexible and may freeze locally.“High-impact constrained option”; add oscillation/freezing tests.More industrially realistic.

§ 07Analysis plan

Primary endpoint

Time to first visible mold by compartment, analysed as time-to-event data. Use Kaplan–Meier curves for “no visible mold yet” and Cox/AFT models for intervention effects. Treat “no failure by end of study” as censored, not as an exact long shelf life.

Model validation metrics

Compare predicted vs observed median failure times using RMSE in days, mean signed bias, accuracy factor, bias factor, and rank correlation across interventions. The most important question is whether the model ranks levers correctly.

Mechanistic calibration

Fit or update nutrient_factor, cocoa L0_mm, rim-drip fraction, package CFU/cm² mapping, and optional biscuit k_pore. Keep a held-out subset for blind prediction.

Safety boundary

Run spoilage validation separately from pathogen safety. For commercial shelf life, a parallel Listeria/RTE safety justification is required; this model should not claim pathogen control.

§ 08Recommended rewritten conclusion for the manuscript

Proposed conclusion The simulator is best presented as a compartmental screening framework for identifying tiramisù spoilage bottlenecks, especially the relative contributions of cocoa, package surfaces, and fill-air deposition. The peer-review comments strengthen the manuscript by shifting emphasis from additive-based preservation toward experimentally testable, legally robust hygiene and ingredient-control levers. Temperature and natamycin remain biologically informative but should be treated as constrained or jurisdiction-dependent interventions rather than default commercial recommendations.

§ 09Key references and regulatory anchors