Integrates networks across different replicates and different species into a single consensus network in a phylogeny-aware manner.
Arguments
- network_list
A named list of
igraphobjects containing the networks of all replicates.- replicate2species
A data frame specifying which species each replicate belongs to, required columns:
- replicate
Character, name of the replicate.
- species
Character, name of the species.
- tree
Object of class
phylo, the phylogenetic tree of the species.
Value
Consensus network in an igraph format with the following edge attributes:
- weight
Numeric, consensus edge weight/adjacency, the weighted average of replicate-wise edge weights.
- n_supporting_replicates
Integer, the number of replicates where the edge was detected (only added if more than 10% of the edges are not present in all replicates).
- supporting_replicates
Character, the list of replicates where the edge was detected (only added if more than 10% of the edges are not present in all replicates).
Details
The input networks should all contain the same nodes (genes). The output consensus network contains the same nodes and all edges that were detected in at least 1 of the replicates.
For each edge, the consensus edge weight (adjacency) is calculated as the weighted mean of replicate-wise edge weights. The weighted mean corrects for 1) the phylogenetic distances between species (if the phylogenetic tree is provided) and 2) the different numbers of replicates per species. As a result, the approach downweighs the edge weights of the replicates that 1) belong to closely related species or 2) belong to species with many replicates, so that an imbalanced sampling across the phylogenetic tree does not bias the consensus network.
If an edge is not present in one of the replicates, the edge weight in that replicate is regarded as 0 for the calculation of the weighted mean. If there are more than 10% such missing/zero-weight edges, the number of replicates and the names of replicate where an edge was detected are saved as 2 new edge attributes ("n_supporting_replicates" and "supporting_replicates", respectively) in the output consensus network.
In the next steps of the pipeline, this consensus network can be used to assign modules jointly for all species while avoiding species bias (see assignInitialModules and pruneModules).