in silico
Phase 1 : In Silico Validation
- Objectif
- Test whether the photoinhibition-repair imbalance model (k_damage > k_NPQ + k_repair) can quantitatively reproduce published LT50 values across Symbiodiniaceae lineages and host genotypes, and identify which parameters (k_de-epox, k_D1, k_synth, Q10) dominate LT50 variance. Kill the hypothesis cheaply if the model cannot fit existing data.
- Coût estimé
- €500-2000
- Durée estimée
- 4-6 weeks
- Critères de réussite
- Model RMSE vs published LT50 · < 1.0°C · (Compare simulated LT50 to Levin et al. 2023 and Baird et al. 2018 values)
- Sobol index for k_de-epox on LT50 · > 0.15 (first-order) · (SALib Sobol analysis)
- Sobol index for k_D1 on LT50 · > 0.15 (first-order) · (SALib Sobol analysis)
- Posterior predictive check · 80% of published LT50 within 95% credible interval · (PyMC posterior predictive sampling)
- On continue si
- Model RMSE < 1.0°C AND k_de-epox and k_D1 each have Sobol index > 0.15 AND posterior predictive check passes 80% coverage
- On arrête si
- Model RMSE > 2.0°C OR neither k_de-epox nor k_D1 has Sobol index > 0.10 (mechanism not dominant) OR model cannot produce LT50 range > 1°C across lineages
- On réoriente si
- Model fits LT50 but sensitivity is dominated by host parameters (k_caspase, k_host_digest) rather than symbiont NPQ/D1 — pivot to host-centric hypothesis
- Risques
- Published LT50 values are confounded by different experimental protocols (ramp rate, duration, light)Probabilité : highParade : Restrict comparison to studies using 48-h ramp at 1°C h⁻¹ with PAR 200-400 µmol m⁻² s⁻¹; run sensitivity analysis on ramp rate
- k_de-epox and k_D1 literature values are from diatoms/plants, not SymbiodiniaceaeProbabilité : highParade : Use wide priors (one order of magnitude) and report posterior uncertainty; flag as key unknown for Phase 2
- Model is over-parameterized and fits any dataProbabilité : mediumParade : Use AICc model comparison against null model (constant LT50) and against simpler 3-parameter model; require ΔAICc > 10 for full model
- No published LT50 for Breviolum B1 or Symbiodinium A3Probabilité : mediumParade : Use RNA-seq thermal tolerance proxies (HSP70 induction temperature) as auxiliary data; flag for Phase 2 measurement