Quickstart¶

Fastest path¶

Run the model with the bundled synthetic stress-test dataset:

python rlto_model.py

This executes example_usage() and will:

evaluate and optimize all demo genes,
compute a global tumor-level optimum,
optimize inhibition against TOXIC_DRIVER,
generate diagnostic plots under diagnostic_plots/,
print a final JSON summary.

Minimal import example¶

from rlto_model import CancerRLTOModel, Microenvironment

env = Microenvironment(
    hypoxia=0.6,
    nutrient_limitation=0.4,
    immune_pressure=0.0,
    drug_pressure=0.0,
    oxidative_stress=0.0,
)

df = CancerRLTOModel.demo_dataset()
model = CancerRLTOModel.from_dataframe(df, microenvironment=env)

results = model.evaluate()
print(results[["gene", "net_fitness", "robustness_probability"]])

Optimize all genes¶

opt_df = model.optimize_all()
print(opt_df[["gene", "baseline_abundance", "optimal_abundance", "optimal_net_fitness"]])

Optimize the global tumor state¶

opt_vector, best_global_fitness = model.optimize_global()
print(best_global_fitness)

Simulate and optimize an intervention¶

intervention_df = model.simulate_intervention("TOXIC_DRIVER", inhibition=0.4)
best_inhibition, residual_fitness = model.optimize_inhibition("TOXIC_DRIVER")

print(best_inhibition)
print(residual_fitness)

Generate diagnostics¶

from rlto_model import ModelDiagnostics

diag = ModelDiagnostics(model)
fig = diag.plot_component_decomposition("MYC")
fig.savefig("MYC_component.png")

Interpreting the first results¶

Focus on these fields first:

net_fitness: the composite objective value per gene,
robustness_probability: analytical survival probability above threshold,
overabundance: mean abundance divided by threshold abundance,
optimal_abundance: abundance maximizing gene-level net fitness.

See Methods and assumptions for the model interpretation.