Heatmap of correlation between latent dimensions — comparison_heatmap

Constructs a heatmap displaying the correlation between the latent dimensions constructed with several integration methods. The lower triangle of the heatmap displays the correlation between features weight, while the upper triangle shows the correlation between samples score. Each triangle of the matrix is reordered separately to show the most highly correlated dimensions next to each other.

Usage

comparison_heatmap_corr(
  output_list,
  latent_dimensions = NULL,
  include_missing_features = FALSE,
  legend_ncol = length(output_list),
  legend_position = "bottom",
  legend_title = "correlation"
)

Arguments

output_list: List of integration methods output each generated via the get_output() function. If named, the names will be used to annotate the plot. See details.
latent_dimensions: Named list, where each element is a character vector giving the latent dimensions to retain in the corresponding element of output_list. Names must match those of output_list. Can be used to filter latent dimensions only in certain elements from output_list (see examples). If NULL (default value), all latent dimensions will be used.
include_missing_features: Logical, see get_features_weight_correlation() for details. Default value is FALSE.
legend_ncol: Integer, number of columns in the legend. Default value is set to the length of output_list.
legend_position: Character, position of the legend. Should be one of "bottom" (default), "top", "left" or "right.
legend_title: Character, name to give to the heatmap colour legend. Intended to use to shorten the legend title if it goes out of frame.

Value

a ComplexHeatmap::Heatmap.

Details

If output_list is unnamed, the different elements in the list will be differentiated by the name of the method used to produce them (e.g. DIABLO, sO2PLS, etc). In order to compare different results from a same integration method (e.g. DIABLO applied to the full vs pre-filtered data), it is possible to assign names to the elements of output_list (see examples). These names will be used in place of the method name in the plot to identify where the latent dimensions come from.

Examples

if (FALSE) {
## Comparing the output from DIABLO, sO2PLS and MOFA

res <- list(
  get_output_diablo(diablo_res), ## diablo_res: output from diablo_run()
  get_output_so2pls(so2pls_res), ## so2pls_res: output from so2pls_o2m()
  get_output_mofa2(mofa_res) ## mofa_res: output from run_mofa
)

comparison_heatmap_corr(res)

## Selecting only some factors from a MOFA run for the comparison
## (for the other methods, all latent dimensions will be retained)
comparison_heatmap_corr(
  res,
  latent_dimensions = list(
    "MOFA" = paste0("Factor ", 1:3)
  )
)

## Comparing two different results from a same integration method -
## diablo_run_full and diablo_run_prefiltered would both be output
## from the diablo_run() function.

res <- list(
  "DIABLO full" = get_output_diablo(diablo_run_full),
  "DIABLO prefiltered" = get_output_diablo(diablo_run_prefiltered)
)

comparison_heatmap_corr(res)
}