impute_test_met: Impute/predict test methylation states

In andreaskapou/Melissa: Bayesian clustering and imputationa of single cell methylomes

Description

Make predictions of missing methylation states, i.e. perfrom imputation using Melissa. This requires keepin a subset of data as a held out test set during Melissa inference. If you want to impute a whole directory containing cells (files) with missing methylation levels, see impute_met_files.

Usage

impute_test_met(
  obj,
  test,
  basis = NULL,
  is_predictive = TRUE,
  return_test = FALSE
)

Arguments

obj	Output of Melissa inference object.
test	Test data to evaluate performance.
basis	Basis object, if NULL we perform imputation using Melissa, otherwise using BPRMeth.
is_predictive	Logical, use predictive distribution for imputation, or choose the cluster label with the highest responsibility.
return_test	Whether or not to return a list with the predictions.

Value

A list containing two vectors, the true methylation state and the predicted/imputed methylation states.

Author(s)

C.A.Kapourani [e-mail]

See Also

create_melissa_data_obj, melissa, filter_regions, eval_imputation_performance, eval_cluster_performance

Examples

# Extract synthetic data
dt <- melissa_synth_dt

# Partition to train and test set
dt <- partition_dataset(dt)

# Create basis object from BPRMeth package
basis_obj <- BPRMeth::create_rbf_object(M = 3)

# Run Melissa
melissa_obj <- melissa(X = dt$met, K = 2, basis = basis_obj, vb_max_iter=10,
   vb_init_nstart = 1, is_parallel = FALSE, is_verbose = FALSE)

imputation_obj <- impute_test_met(obj = melissa_obj,
                                   test = dt$met_test)

andreaskapou/Melissa documentation built on June 12, 2020, 5:54 p.m.