in silico
Phase 1 : In Silico Validation
- Objectif
- Rapidly test whether the causal chain (fiber/probiotic -> systemic SCFA -> intra-ovarian SCFA -> HDAC inhibition + GPR41/43 activation -> reduced ROS/NF-kB -> improved follicle/oocyte outcomes) is internally coherent and supported by existing transcriptomic, proteomic and clinical data, and pre-register the falsifiable predictions with calibrated effect sizes before any animal is used.
- Coût estimé
- €0-2000
- Durée estimée
- 4-6 weeks
- Critères de réussite
- Detectable co-expression of FFAR2 or FFAR3 and HDAC1-3 in granulosa/immune ovarian cells · FFAR2/FFAR3 detected in >=10% of granulosa cells and HDAC1-3 in >=30% in at least one dataset · (scRNA-seq pseudobulk + fraction expressing)
- Model reproduces published butyrate->H3K9ac fold change · Simulated H3K9ac fold change within 2-3x of the 10.1096/fj.202000444R reported value · (ODE calibration against extracted data points)
- Power feasibility for prediction 1 · n<=12/group achieves 80% power at alpha=0.05 for >=30% serum butyrate increase · (simr simulation with extracted variance)
- Physiological dose window identified · At least one published mouse butyrate dose yielding serum <100 uM with measurable ovarian effect · (PK data extraction and dose interpolation)
- On continue si
- FFAR2/FFAR3 and HDAC co-expression confirmed AND model identifies a dominant testable node AND power feasible with n<=15/group for predictions 1-4
- On arrête si
- No detectable FFAR2/FFAR3 in ovarian tissue across all datasets OR model cannot reproduce published butyrate effects within 3x (mechanism implausible)
- On réoriente si
- GPR signal absent but HDAC arm robust -> refocus hypothesis on HDAC-only mechanism and drop Ffar KO arm in Phase 2/3
- Risques
- Public ovarian scRNA-seq datasets too sparse for immune-cell FFAR detectionProbabilité : mediumParade : Supplement with bulk deconvolution (CIBERSORTx) and targeted literature on ovarian macrophage GPR43
- Published effect sizes heterogeneous, inflating/deflating powerProbabilité : highParade : Use conservative (lower bound) effect size and inflate n by 20%; pre-specify sensitivity analysis
- Model overfitting to sparse kinetic dataProbabilité : mediumParade : Report parameter identifiability; treat model as hypothesis-ranking tool, not predictor