Page 1 of 7
Historical data
Start from the historical runs a customer would already have.
Page 1 data gathered
Customer data starts as runs, factors, and measured outcomes
The synthetic dataset has 28 historical fermentation runs with strain, medium, feed, pH, temperature, DO, and final process readouts.
Measured run trajectory
Selected run facts
Page 2 process metrics
Raw measurements become decision metrics
Titer alone is not enough. The demo calculates yield, productivity, carbon efficiency, byproduct burden, viability proxy, and stress penalties before ranking.
Titer versus carbon efficiency
Screen-only ranking
Page 3 bottleneck analysis
Factor effects show where the next experiment should focus
This is intentionally practical: enough statistics to decide what to test next, not a heavy academic DOE interface.
ANOVA-style factor signal
Observed bottlenecks
Factor response map
Page 4 model review
Process data is translated into FBA-style constraints
The model layer checks whether the measured phenotype is feasible, which constraints are strained, and which high-titer runs are risky.
Selected run constraints
Model-adjusted ranking
Page 5 next DOE
The output is a small next-run plan, not a giant factorial
The system recommends 8 runs that test the main bottlenecks while staying close to feasible operating space.
Page 6 DOE results
Concordance is the proof step
The system does not stop at suggesting DOE runs. Once follow-up results return, it compares predicted outcomes with observed titer, yield, and byproduct burden.
Predicted versus observed titer
Result interpretation
Page 7 partner memo
Prospect-ready output
The final artifact is a concise recommendation package: what happened, what seems limiting, and what the next wet-lab run should test.
Assumptions and limitations
This demo uses synthetic data and a simplified model-review layer. It is meant to show the customer workflow and decision logic, not claim exact intracellular flux prediction. In a real trial, the model is configured from the customer's organism, GEM/SBML files, medium composition, historical runs, and assay context.