Plots features weight against covariate — plot_features_weight

Plots the features weight or importance from the result of an integration method against a covariate from the features metadata.

Usage

plot_features_weight_covariate(
  method_output,
  mo_data,
  covariate,
  features_metric = c("signed_importance", "weight", "importance"),
  remove_null_weight = FALSE,
  latent_dimensions = NULL,
  colour_by = NULL,
  shape_by = NULL,
  point_alpha = 0.5,
  add_se = TRUE,
  add_boxplot = TRUE,
  scales = "free_x"
)

Arguments

method_output: Integration method output generated via the get_output() function.
mo_data: A MultiDataSet object (will be used to extract samples information).
covariate: Character or named list of character, giving for each dataset the name of the column in the corresponding features metadata to use as x-axis in the plot. If one value, will be used for all datasets. If list, the names must correspond to the names of the datasets in mo_data. If a dataset is not present in this list, will be excluded from the plot.
features_metric: Character, the features metric that should be plotted on the y-axis. Should be one of 'signed_importance' (default value), 'weight' or 'importance'.
remove_null_weight: Logical, should features with null weight/importance be removed from the plot? Default value is FALSE.
latent_dimensions: Character vector giving the latent dimensions to display. Default value is NULL, i.e. all latent dimensions will be shown.
colour_by: Character or named list of character, giving for each dataset the name of column in the corresponding feature metadata to use to colour the features in the plot. If one value, will be used for all datasets. If list, the names must correspond to the names of the datasets in covariate. Default value is NULL.
shape_by: Character or named list of character, giving for each dataset the name of column in the corresponding feature metadata to use as shape for the features in the plot. If one value, will be used for all datasets. If list, the names must correspond to the names of the datasets in covariate. Default value is NULL.
point_alpha: Numeric between 0 and 1, the opacity of the points in the plot (with 1 = fully opaque, and 0 = fully transparent). Default value is 0.5.
add_se: Logical, should a confidence interval be drawn around the smoothing curves for numerical covariates? Default value is TRUE.
add_boxplot: Logical, should a boxplot be drawn on top of the points for categorical covariates? Default value is TRUE.
scales: Character, value to use for the scales argument of ggplot2::facet_grid(). Default value is 'free_x'.

Details

If the covariate is numeric, the function creates a scatter plot, with a loess curve to summarise the trend between the covariate and the features weight. If colour_by is used, and the corresponding variable is numeric, the loess curve will not take into account this variable. If instead the colour_by variable is a character or factor, a loess curve will be fitted separately for each category.

If the covariate is not numeric, the function creates a violin/boxplot. If colour_by is used, and the corresponding variable is numeric, the violins and boxplots will not take into account this variable. If instead the colour_by variable is a character or factor, a separate violin and boxplot will be drawn for each category.