robustnet.sampling

Define class for sampling model parameters.

Attributes

logger

Classes

Sampler

SamplingCallback

Module Contents

robustnet.sampling.logger[source]
class robustnet.sampling.Sampler(model, exclude_metabs=None)[source]

Bases: robustnet.kinetics.Fitter

_make_value_array(data, varnames, argname, method='geomean')[source]

Make an array with elements ordered according to varnames.

If data is a dict or pandas.Series, missing values (either absent or NaN) are imputed by method.

_prepare_prior(mu, sigma, init, name, label)[source]

Prepare prior mu, sigma and initvalues for variables used in sampling.

static log_transform(log_mu, log_sigma)[source]

Assume X ~ LogNormal(log_mu, log_sigma). Compute the mu and sigma of the corresponding normal distribution: log(X) ~ Normal(mu, sigma).

_to_ser(data, name)[source]

Convert a numpy.ndarray to a pandas.Series

sample_with_omics(ref_v_prior_mu, kparam_prior_mu, x_prior_mu, e_prior_mu, ref_v_prior_sigma=0.01, kparam_prior_sigma=0.1, x_prior_sigma=0.1, e_prior_sigma=0.001, ref_v_initvalues=None, kparam_initvalues=None, x_initvalues=None, e_initvalues=None, alpha=None, n_tunes=10000, n_samples=10000, n_chains=10, n_jobs=1)[source]

Sample model parameters with fluxomics, metabolomics, proteomics data and enzyme kinetic parameters.

Parameters:

  • ref_v_prior_mu (dict or pandas.Series) – Reference-state flux distribution in units of mmol/L/s (cell-based). If None, reference fluxes loaded by load_priors are used.

  • kparam_prior_mu (dict or pandas.Series) – Mean values of the prior distributions for kinetic parameters. Catalytic constants have units of 1/s, Michaelis, activation, and inhibition constants have units of mM, and equilibrium constants are dimensionless. Missing kinetic parameters are allowed. If None, kinetic parameters loaded by load_priors are used.

  • x_prior_mu (dict, pandas.Series or None) – Mean values of prior metabolite concentrations in mM (cell-based). Missing metabolites are allowed. If None, metabolomics data loaded by load_priors are used.

  • e_prior_mu (dict, pandas.Series or None) – Mean values of prior enzyme concentrations in mM (cell-based). Missing enzymes are allowed. If None, proteomics data loaded by load_priors are used.

  • ref_v_prior_sigma (scalar, dict, pandas.Series, optional) – Standard deviations of reference-state fluxes. If a scalar is provided, the same value is used for all fluxes. Missing flux values are allowed when using a dict or pandas.Series. If None, standard deviations from reference fluxes loaded by load_priors are used. Defaults to 0.01.

  • kparam_prior_sigma (scalar, dict or pandas.Series, optional) – Standard deviations of kinetic parameters. If a scalar is provided, the same value is used for all parameters. Missing parameter values are allowed. If None, standard deviations from kinetic parameters loaded by load_priors are used. Default to 0.1.

  • x_prior_sigma (scalar, dict or pandas.Series or None, optional) – Standard deviations of metabolite concentrations. If a scalar is provided, the same value is used for all metabolites. Missing metabolite values are allowed. If None, standard deviations from metabolomics data loaded by load_priors are used. Default to 0.1.

  • e_prior_sigma (scalar, dict or pandas.Series or None, optional) – Standard deviations of enzyme concentrations. If a scalar is provided, the same value is used for all enzymes. Missing enzyme values are allowed. If None, standard deviations from proteomics data loaded by load_priors are used. Default to 0.001.

  • ref_v_initvalues (dict, pandas.Series or None, optional) – Initial values for reference flux sampling. Missing fluxes are allowed. If None, ref_v_prior_mu is used.

  • kparam_initvalues (dict, pandas.Series or None, optional) – Initial values for kinetic parameter sampling. Missing parameter values are allowed. If None, kparam_prior_mu is used.

  • x_initvalues (dict, pandas.Series or None, optional) – Initial values for metabolite concentration sampling. Missing metabolite values are allowed. If None, x_prior_mu is used.

  • e_initvalues (dict, pandas.Series or None, optional) – Initial values for enzyme concentration sampling. Missing enzyme values are allowed. If None, e_prior_mu is used.

  • alpha (float or None, optional) –

    Gaussian penalty strength used in parameter balancing. Larger values impose stronger penalties in log-posterior space.

    A reasonable choice is often on the same order of magnitude as 1 / ref_flux_sigma**2. If None, geomean(1 / ref_flux_sigma**2) is used.

  • n_tunes (int, optional) – Number of tuning iterations performed before sampling in each chain.

  • n_samples (int, optional) – Number of samples drawn in each chain.

  • n_chains (int, optional) – Number of sampling chains.

  • n_jobs (int, optional) – Number of parallel jobs to run in parallel.

class robustnet.sampling.SamplingCallback(pbar)[source]
pbar[source]
traces[source]
__call__(trace, draw)[source]