Configuration Reference

U-MIMIC uses YAML configuration files validated by Pydantic. All fields have sensible defaults, so you only need to specify what you want to change.

Top-Level: ExperimentConfig

Field	Type	Default	Description
name	str	"experiment"	Experiment name
context	str	"in_vitro"	"in_vitro" or "in_vivo"
seed	int	42	Random seed for reproducibility
dynamics	DynamicsConfig	see below	Cell-state dynamics settings
pk	PKConfig	see below	Pharmacokinetic model settings
dosing	DosingConfig	see below	Dosing schedule settings
observations	ObservationConfig	see below	Observation model settings
inference	InferenceConfig	see below	Inference engine settings
priors	PriorConfig	see below	Prior distribution settings
data	DataConfig	see below	Data loading settings
simulation	SimulationConfig	see below	Simulation / synthetic data settings

DynamicsConfig

Field	Type	Default	Description
states	list[str]	["P","Q"]	Active cell states (P, Q, A, R)
density_dependent	bool	false	Enable logistic density dependence
carrying_capacity	`float	null`	null
clearance_rate	float	0.1	Apoptotic cell clearance rate (1/h)

PKConfig

Field	Type	Default	Description
model	str	"none"	"none", "one_compartment", "two_compartment"
vd	float	10.0	Volume of distribution (1-compartment)
ke	float	0.1	Elimination rate constant (1/h)
ka	float	null	Absorption rate constant (oral dosing)
vc	float	null	Central volume (2-compartment)
vp	float	null	Peripheral volume (2-compartment)
cl	float	null	Clearance (2-compartment)
q	float	null	Inter-compartmental clearance

DosingConfig

Field	Type	Default	Description
type	str	"constant"	"constant", "single_bolus", "repeated_bolus", "oral"
concentrations	list[float]	null	In-vitro concentration list
dose_amount	float	null	Dose amount per administration
interval	float	null	Dosing interval (hours)
n_doses	int	null	Number of doses
start_time	float	0.0	Time of first dose (hours)

ObservationConfig

Field	Type	Default	Description
modalities	list[str]	["cell_counts"]	Enabled modalities
cell_count_overdispersion	float	10.0	Negative Binomial overdispersion
bli_alpha	float	1000.0	BLI photons-per-cell scaling
bli_sigma_log	float	0.3	BLI log-normal noise sigma
volume_beta	float	1e-3	Volume conversion factor
volume_sigma	float	0.2	Volume measurement noise
biomarker_precision	float	50.0	Biomarker precision parameter

InferenceConfig

Field	Type	Default	Description
mode	str	"mle"	"mle", "mcmc", "smc", "hierarchical"
backend	str	"scipy"	"scipy", "emcee", "pymc", "particle"
n_samples	int	2000	MCMC samples per chain
n_chains	int	4	Number of MCMC chains
n_warmup	int	1000	MCMC warmup / burn-in samples
n_particles	int	500	SMC particle count
n_restarts	int	5	MLE multi-start restarts

SimulationConfig

Field	Type	Default	Description
method	str	"gillespie"	"ode", "gillespie", "tau_leaping"
initial_cells	int	100	Initial cell count
t_max	float	72.0	Simulation duration (hours)
dt_obs	float	4.0	Observation interval (hours)
n_replicates	int	4	Replicates per condition
seed	int	42	Simulation random seed

PriorConfig

Prior distributions for inference parameters. Each entry is a dict of distribution parameters (keys depend on the distribution family).

Field	Default	Description
b0	{scale: 0.04, s: 0.5}	Birth rate prior
d0_P	{scale: 0.01, s: 0.5}	Death rate (P) prior
emax_death	{scale: 0.1}	Emax death prior
ec50_death	{scale: 1.0, s: 1.0}	EC50 death prior
hill_death	{scale: 1.5, s: 0.3}	Hill coefficient prior

Complete Example

name: cytotoxic_breastcancer
context: in_vitro
seed: 123

dynamics:
  states: [P, Q]
  density_dependent: false

dosing:
  type: constant
  concentrations: [0, 0.01, 0.03, 0.1, 0.3, 1.0, 3.0, 10.0]

observations:
  modalities: [cell_counts]
  cell_count_overdispersion: 15.0

simulation:
  method: gillespie
  initial_cells: 200
  t_max: 96.0
  dt_obs: 6.0
  n_replicates: 3

inference:
  mode: mle
  backend: scipy
  n_restarts: 10