T01 — Simulate a Training Experiment¶

Goal: Generate a 27-run Box-Behnken training dataset using the CHO CloneX simulator and produce Fig. 4 (training trajectories).

Script: examples/01_simulate_training_experiment.py

What the script does¶

Instantiates CHOSimulator(clone="CloneX", seed=0).
Calls generate_box_behnken_experiment(n_days=28, seed=0), which returns 27 runs (4-factor Box-Behnken design with 3 centre points).
Serialises each run to runs/run_NNN.json.
Plots VCD trajectories and saves fig4_training_trajectories.pdf.

Running¶

python examples/01_simulate_training_experiment.py
# → runs/run_000.json … run_026.json
# → fig4_training_trajectories.pdf

Alternatively, via the CLI:

perfusio simulate --clone CloneX --n-days 28 --out runs/

Key code¶

from perfusio.simulator.cho_perfusion import CHOSimulator
from perfusio.viz.static import fig4_training_trajectories

sim = CHOSimulator(clone="CloneX", seed=0)
runs = sim.generate_box_behnken_experiment(n_days=28, seed=0)
# len(runs) == 27  (24 edge + 3 centre)

fig = fig4_training_trajectories(runs, species="VCD")
fig.savefig("fig4.pdf")

Design space¶

The four factors and their ranges used in the Box-Behnken design:

Factor	Low	Centre	High	Units
Perfusion rate	0.5	1.0	1.5	vvd
Bleed rate	0.05	0.10	0.15	vvd
Glucose setpoint	3.0	4.5	6.0	g L⁻¹
Temperature	36.0	36.5	37.0	°C

Expected output¶

VCD reaches ~8–12 × 10⁶ cells mL⁻¹ at steady state under optimal conditions.
Titer accumulates to ~400–600 mg L⁻¹ by day 28 for high-performing runs.

Next step¶

Proceed to T02 — Fit the Hybrid GP Model.