impute_met_state {Melissa} | R Documentation |
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.
impute_met_state(obj, test, basis = NULL, is_predictive = TRUE, return_test = FALSE)
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. |
A list containing two vectors, the true methylation state and the predicted/imputed methylation states.
C.A.Kapourani C.A.Kapourani@ed.ac.uk
create_melissa_data_obj
, melissa
,
filter_regions
, eval_imputation_performance
,
eval_cluster_performance
# 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$met, K = 2, basis = basis_obj, vb_max_iter = 10, vb_init_nstart = 1, is_parallel = FALSE, is_verbose = FALSE) imputation_obj <- impute_met_state(obj = melissa_obj, test = dt$met_test)