Worked examples

Every output below was reproduced on a clean install of release stable 6588707. All of it is model output, and none of it is qualified (NOT_QUALIFIED).

What can and cannot run

livemore programs list

Six programs, and each one tells you its own state:

Program	Material	State
Exogenous insulin	rapid-acting analogue (SC), regular human insulin (IV)	runs
LBHR-030 Greater burdock	arctigenin	margin computable
LBHR-135 Bilberry	cyanidin-3-glucoside	runs at the studied dose only
LBHR-172 Lion's mane	erinacine A	exposure only
LBHR-129 Yerba santa	sterubin	blocked, reason named
LBHR-062 Rosemary	carnosic acid	blocked, reason named

See reference programs for what each state means.

Paired insulin arms across 12 research members

livemore programs run insulin --cohort 12 --seed 0 --json

Twelve research members, each with their own paired control and treated arm from an identical starting state:

"summary": {
  "member_count": 12,
  "mean_delta_glucose_auc_mmol_L_min": -1700.421,
  "delta_glucose_auc_range": [-2975.87, -937.221],
  "members_with_any_time_below_3_9_treated": 1
},
"evidence_state": "MODELED_ONLY",
"qualification_state": "NOT_QUALIFIED",
"applicability": "Insulin-deficient physiology only: the model has no endogenous insulin secretion. Not applicable to preserved beta-cell function."

One of twelve members went below 3.9 mmol/L, the hypoglycaemia threshold. The result reports that alongside the mean glucose-AUC reduction rather than reporting only the average benefit.

A refusal that names its own blocker

$ livemore programs run LBHR-135 --dose-mg 100
  run=protocol-44850896… · program=LBHR-135 · outcome=EXPOSURE_NOT_COMPUTABLE · reason=DOSE_NOT_MEASURED_AT_THIS_LEVEL · exposure_basis=modelled · human_exposure_admissible=True · controls_passed=True · persisted=False
  missing: Measured human exposure exists only for the 500 mg dose in the cited study. Requesting 100 mg would require scaling a measured peak concentration through a kinetic model, which this program does not have.

Now ask the question the evidence can answer:

$ livemore programs run LBHR-135 --dose-mg 500
  run=protocol-e1dfc548… · program=LBHR-135 · outcome=REACHES_MEASURED_ACTIVE_RANGE · summary={'unbound_cmax_um': 0.141, 'margin_to_active_low': 1.41, 'margin_at_one_standard_error_below_cmax': 0.71, 'robust_to_one_standard_error': False, 'apparent_clearance_L_per_h': 3989.6} · exposure_basis=measured_human_summary_statistics · human_exposure_admissible=True · controls_passed=True · persisted=False

At 500 mg, the modeled unbound peak concentration reaches the measured active range, with a margin of 1.41. One standard error lower, the margin falls to 0.71, below the range. The answer is "yes, but not robustly", and the software states both halves. This exposure margin comes from summary statistics of a small published human study and does not establish efficacy.

Genome-scale metabolism under restricted oxygen

This solves Human-GEM 2.0, a genome-scale model of human metabolism with 12,931 reactions, 8,461 metabolites and 2,848 genes. It asks how much ATP the network can make when oxygen is restricted but glucose is not.

curl -fsS -X POST -H "Authorization: Bearer $(livemore token)" -H 'Content-Type: application/json' \
  -d '{"conditions":[
        {"condition_id":"normoxia","glucose_uptake":1.0,"oxygen_uptake":1.0},
        {"condition_id":"hypoxia","glucose_uptake":1.0,"oxygen_uptake":0.15}]}' \
  http://127.0.0.1:8852/protocols/human-gem-oxygen-capacity

Condition	ATP capacity (mmol/gDW/h)	Oxygen	Glucose	Mass-balance residual
normoxia	7.00	1.00	1.00	5 × 10⁻¹⁵
hypoxia	2.75	0.15	1.00	9 × 10⁻¹⁶

With the same glucose, restricting oxygen lowers the network's maximum ATP-maintenance capacity by about 60%. This is a capacity bound. The result does not claim a unique flux distribution, and it states its own limits:

"claims_allowed":    ["conditional stoichiometric ATP capacity", "numerical conservation and constraint checks"],
"claims_prohibited": ["measured human response", "ATP concentration", "unique flux prediction",
                      "drug efficacy", "patient-specific physiology", "clinical validation"]

Open any result in Jupyter

livemore notebook

Any run can also be downloaded as analysis.csv and analysis.ipynb. See Jupyter and exports. The notebook reproduces the recorded result rather than recomputing it.