[INTERNAL] A wrapper functions that calls the functions to generate a network from
correlation data and reduce the network by a given method. Correlation/adjacency matrices are
computed in compute_correlation_matrices. Graph generation uses
graph.adjacency internally. Methods implemented are
network_reduction_by_p_value (reduction by statistical significance of correlation)
and network_reduction_by_pickHardThreshold (using WGCNA function
pickHardThreshold.fromSimilarity that finds a suitable cutoff value to get a scale-free
network). If no method is given, no reduction will be performed. When using the reduction method
`p_value` the user can specify an alpha significance value and a method for p-value adjustment.
When using the reduction by `pickHardThreshold` a R^2 cutoff and a cut vector can be specified.
generate_reduced_graph(
adjacency_matrix,
measurement_data,
identifiers,
handling_missing_data = "all.obs",
reduction_method = "pickHardTreshold",
r_squared_cutoff = 0.85,
cut_vector = seq(0.2, 0.8, by = 0.01),
mean_number_edges = NULL,
edge_density = NULL,
p_value_adjustment_method = "BH",
reduction_alpha = 0.05,
n_threads = 1,
parallel_chunk_size = 10^6,
print_graph_info = TRUE
)iGraph graph object of the reduced network.
[matrix] Adjacency matrix of correlations computed using cor in
compute_correlation_matrices
[data.frame] Data frame containing the respective raw data (e.g. mRNA expression data, protein abundance, etc.) to the adjacency matrix. Analyzed components (e.g. genes) in rows, samples (e.g. patients) in columns.
[data.frame] Data frame containing biological identifiers and the corresponding node ID
created in compute_correlation_matrices via create_unique_layer_node_ids.
The column containing node IDs has to be named `node_id`.
["all.obs"|"pairwise.complete.obs"] Specifying the handling of missing data during correlation matrix computation. (default: all.obs)
["pickHardThreshold"|"p_value"] A character string specifying the method to be used for network
reduction. `p_value` for hard thresholding based on the statistical significance of the
computed correlation. `pickHardThreshold` for a cutoff based on the scale-freeness criterion
(calls pickHardThreshold). (default: pickHardThreshold)
[float] A number indicating the desired minimum scale free topology fitting index R^2 for reduction
using pickHardThreshold. (default: 0.85)
[sequence of float] A vector of hard threshold cuts for which the scale free topology fit indices are to
be calculated during reduction with pickHardThreshold. (default: seq(0.2, 0.8, by = 0.01))
[int] Find a suitable edge weight cutoff employing pickHardThreshold to reduce
the network to at most the specified mean number of edges. Attention: This parameter overwrites the 'r_squared_cutoff' and
'edge_density' parameters if not set to NULL. (default: NULL)
[float] Find a suitable edge weight cutoff employing pickHardThreshold to reduce the
network to at most the specified edge density. Attention: This parameter overwrites the 'r_squared_cutoff' parameter if not set
to NULL. (default: NULL)
["holm"|"hochberg"|"hommel"|"bonferroni"|"BH"|"BY"|"fdr"|"none"] String of the correction method applied to p-values. Passed to p.adjust. (default: "BH")
[float] A number indicating the significance value for correlation p-values during reduction. Not-significant edges are dropped. (default: 0.05)
[int] Number of threads for parallel computation of p-values during p-value reduction. (default: 1)
[int] Number of p-values in smallest work unit when computing in parallel during network reduction with method `p_value`. (default: 10^6)
[bool] Specifying if a summary of the reduced graph should be printed to the console after network generation. (default: TRUE)