DOE model workflow

Start with raw data.Then choose the nextexperiment set for money.

Six steps: show the raw media data, fit a simple ANOVA model, check the FBA model, improve FBA concordance, rank measured conditions, then simulate new nearby candidates.

Input data first. 16 formulation conditions, 6 media components, measured biomass, cost estimates, FBA flux outputs, and model-fit tables. The demo keeps raw measurements, statistical predictions, and FBA predictions separate.

Page 1 of 6

Raw data

Show the exact formulation data before any model output.

Step 1 raw data

Show the raw data first

This page shows the full dataset before any model, score, or recommendation.

Raw formulation table

All 16 formulation conditions are shown here. Component values are the media amounts from the supplied dataset; biomass and cost are measured or estimated outputs; biomass flux is model-derived.

Measured rows16 formulation conditions
Measured columns6 media components + biomass
Derived model datamolarities, fluxes, fit metrics
Next output5 cost-aware tests

Measured biomass

Cost per gram biomass

Step 2 ANOVA model analysis

ANOVA asks which media components explain biomass

This page fits an ANOVA-style main-effects model from media components to measured biomass, then reports R² values.

Measured versus ANOVA-predicted biomass

R² reports how much of the biomass variation is explained by the fitted main-effects model.

ANOVA factor effects

Selected condition

ANOVA prediction table

Measured biomass compared with the simple ANOVA main-effect prediction.

Step 3 FBA analysis

FBA is checked against the same measured biomass data

This page shows the baseline FBA model concordance. It uses the model-derived flux signal and compares it with measured biomass.

Measured biomass versus FBA prediction

Model signal is scaled for visual comparison; the reported concordance is the audit metric.

Largest visible gaps

Step 4 improve FBA concordance

Use the FBA mismatch to improve model concordance

The model update is judged by one thing: did concordance with measured biomass improve?

Improved FBA concordance

Model updates that improved fit

Technical hypothesis table

Step 5 next set

Rank measured conditions for money

These are previously measured formulations, not new predictions. The next page explores new nearby conditions.

Measured reference formulations

Money objective ranking

Cost and FBA signals used

Flux evidence used in the cost review

Step 6 Monte Carlo

Explore the next experiments

Generate new formulations near the measured conditions. Bootstrap models estimate a range of possible outcomes; nothing on this page is a measured result or a new FBA run.

Chance of meeting both targets

Predicted biomass vs cost

Five suggested next tests

All simulated formulations

Exploratory only: 16 observations, six factors, local candidate changes, ridge-regression proxies, and bootstrap uncertainty. Cost is estimated from the recorded formulation costs. Confirm feasibility and remeasure selected candidates before drawing process conclusions.