meMultiplex.RdConstructs a multiplex network from two `igraph` layers and writes the multiplex structure to file. Layers should share some, but not necessary all, nodes. Function interprets the first layer as correlation layer. The second layer, if weighted, should have only non.negative weights and can be interpreted as distance layer. Multiplex structure can be written as `Infomap` of `multinet` object.
meMultiplex( cor_layer, supplementary_layer, output_file, cor_weighted = TRUE, supp_weighted = TRUE, cor_normalization_fun = meNet::max_normalization, supp_normalization_fun = meNet::neg_max_normalization, output_type = "infomap", inter_cor_supp = NULL, inter_supp_cor = inter_cor_supp )
| cor_layer | Correlation layer of the multiplex as `igraph` object. |
|---|---|
| supplementary_layer | Supplementary layer of the multiples as `igraph` object. |
| output_file | Name of the output file to which multiplex structure is written. |
| cor_weighted | Whether the correlation layer is weighted. If `FALSE`, all weights are set to `1`. Defaults to `TRUE`. |
| supp_weighted | Whether the supplementary layer is weighted. If `FALSE`, all weights are set to `1`. Defaults to `TRUE`. |
| cor_normalization_fun | Normalization function applied on the weights of the correlation layer, if the layer is weighted. If `NULL`, no normalization is applied. Defaults to `max_normalization`. |
| supp_normalization_fun | Normalization function applied on the weights of the supplementary layer, if the layer is weighted. If `NULL`, no normalization is applied. Defaults to `neg_max_normalization`. |
| output_type | Structure of the output file. Has to be one of the values "infomap"`` or `"multinet"`. Default value is `"infomap"`. |
| inter_cor_supp | Weight of the inter-layer edges from the correlation layer to the supplementary layer. Defaults to `NULL`. |
| inter_supp_cor | Weight of the inter-layer edges from the supplementary layer to the correlation layer. By default, the values is equal to `inter_cor_supp`. |
A vector of removed nodes. Invisibly, structure of the created multiplex is saved to the `output_file`.
Layers could previously be constructed using some of the functions `meNet_cor`, `meNet_CGI` or `meNet_gene`. If layers are weighted, function searches for the edge attributes with names`"Cor"` and `"Dist"`. If the attribute doesn't exist, it takes the first edge attribute instead. The function handles weight of the first layer as correlations and weights of the second layer as distances.
Function checks if all edge weights of `supplementary_layer` are positive. For negative weights of `correlation_layer`, corresponding edge is removed from both layers. In this step more power is given to significantly negative correlations in community breakage. After the removal of negative edges, nodes which are isolated in both layers are removed from the multiplex.
Additional parameters can be set, such as the weight of inter-layer edges. `inter_cor_supp` defines the weight of all inter-layer edges from the correlation layer to the supplementary layer, while `inter_supp_cor` defines the weight of all inter-layer edges from the supplementary layer to the correlation layer. By default, both values are `NULL`. If `NULL`, for `Infomap` output inter-layer edges are not added to file while for `multinet` output the type of network is defined as `multiplex` instead of `multilayer`.
If a layer is weighted, normalization function can be used on the edge weights. For correlation layer the default is maximum normalization using `max_normalization` function. For supplementary layer, the default is negative maximum normalization with `neg_max_normalization`. The difference in normalization function exists because weights of correlation layer can be interpreted as "strengths" while the weights of supplementary layer can be interpreted as "cost". That is, large correlation has opposite impact than the large distance for a pair of nodes.
Resulting multiplex structure is written to a file either in the `Infomap` (De Domenico et al. 2015) format or in the `multinet` (Magnani et al. 2021) format, depending on the `output_type`. `Infomap`-style file can be used by `Infomap` algorithm online or from the command line upon installation. `multinet`-style file can be converted to `multinet` object using `meMultiplex_to_multinet` function.
De Domenico M, Lancichinetti A, Arenas A, Rosvall M (2015).
“Identifying modular flows on multilayer networks reveals highly overlapping organization in interconnected systems.”
Physical Review X, 5(1), 011027.
Magnani M, Rossi L, Vega D (2021).
“Analysis of Multiplex Social Networks with R.”
Journal of Statistical Software, 98(1), 1--30.