ml_corr_network — ml_corr_network • LipidSigR

Select feature importance from the result of ml_model. Two methods are available to rank feature importance: algorithm-based and SHAP analysis. The average importance of the top features across all CV runs will be displayed by specifying a feature count.

ml_corr_network(
  ml_se,
  feature_importance = "Algorithm-based",
  correlation = "pearson",
  edge_cutoff = 0,
  feature_num = 10,
  nsim = 5
)

Arguments

ml_se: A SummarizedExperiment object with results computed by ml_model.
feature_importance: Character. The method of feature importance. Allowed methods are 'Algorithm-based' and 'SHAP'. Default is 'Algorithm-based'.
correlation: Character. The method for computing correlation coefficient. Allowed methods includes 'pearson', 'kendall', and 'spearman'. Default is 'pearson'.
edge_cutoff: Numeric. A value between 0 and 1. Only the correlation coefficient larger than it will be shown as a line in the plot.
feature_num: Numeric. The number of features to be shown in the plots. A feature number value selected from feature_option of the SummarizedExperiment object returned by ml_model. Usually be one of 2, 3, 5, 10, 20, 50, 100. Default is 10.
nsim: Integer. The times of simulation. Default is 5.

Value

Return 1 interactive plot, 1 static plot, and 1 table.

interactive_correlation_network & static_correlation_network: correlation network.
edge_table & node_table: table for plotting network.

Examples

data("ml_se_sub")
res <- ml_corr_network(ml_se_sub, feature_importance='Algorithm-based',
    correlation='pearson', edge_cutoff=0, feature_num=10, nsim=5)