R6 class for frozen batch effect correction using ComBat algorithm.
Parameters are estimated on training data and can be applied to new data
without re-estimation, preventing data leakage.
Methods
Method new()
Create a new FrozenComBat object
Arguments
parametricLogical, use parametric priors (default TRUE)
mean_onlyLogical, only adjust means, not variances (default FALSE)
Returns
A FrozenComBat object
Method fit()
Fit ComBat parameters on training data
Usage
FrozenComBat$fit(data, batch, covariates = NULL)
Arguments
dataNumeric matrix with features in columns, samples in rows
batchFactor or character vector indicating batch membership
covariatesOptional data.frame of biological covariates to preserve
Returns
Self (invisibly), for method chaining
Apply frozen ComBat correction to data
Usage
FrozenComBat$transform(data, batch)
Arguments
dataNumeric matrix with features in columns, samples in rows
batchFactor or character vector indicating batch membership
Returns
Batch-corrected matrix with same dimensions as input
Fit and transform in one step (convenience method)
Usage
FrozenComBat$fit_transform(data, batch, covariates = NULL)
Arguments
dataNumeric matrix
batchBatch vector
covariatesOptional covariates
Method is_fitted()
Check if the object has been fitted
Method get_batch_levels()
Get the batch levels learned during fitting
Usage
FrozenComBat$get_batch_levels()
Returns
Character vector of batch level names
Print method
Estimate ComBat parameters from training data
This implements the core empirical Bayes estimation from ComBat
Apply frozen correction parameters to new data
Method clone()
The objects of this class are cloneable with this method.
Usage
FrozenComBat$clone(deep = FALSE)
Arguments
deepWhether to make a deep clone.