Extracts the feature weights from a trained MOFA or MEFISTO model (from the MOFA2 package).
Usage
mofa_get_weights(
object,
views = "all",
factors = "all",
abs = FALSE,
scale = "none",
as.data.frame = TRUE
)Arguments
- object
A trained MOFA object.
- views
Character or integer vector, name or index of the views (i.e. datasets) for which the feature weights should be extracted. Default value is "all", i.e. all datasets are considered.
- factors
Character or integer vector, name or index of the factors for which the feature weights should be extracted. Default value is "all", i.e. all factors are considered.
- abs
Logical, should the absolute value of the weights be returned? Default value is
FALSE.- scale
Character, the type of scaling that should be performed on the feature weights. Possible values are
'none','by_view','by_factor'or'overall'(see Details). Default value is'none'.- as.data.frame
Logical, whether the function should return a long data-frame instead of a list of matrices. Default value is
TRUE.
Value
By default, returns a tibble with columns view, feature, factor, value. Alternatively, if
as.data.frame = FALSE, returns a list of matrices, one per view, with features as rows and factors as columns.
Details
Scaling options:
scale = 'none': no scaling performed;scale = 'by_view': weights are divided by the maximum absolute weight in the corresponding view/dataset;scale = 'by_factor': weights are divided by the maximum absolute weight in the corresponding factor;scale = 'overall': weights are divided by the maximum absolute weight across all views/datasets and factors considered.