
Identify hub TFs
identify_hub_TFs.RdIdentifies hub TFs per module using values of module membership and degree.
TFs with high module membership (correlation of their activity with the module score > mm_thresh, with the
correlation used to build the network, corr_type of compute.WTCNA())
and a degree at or above the degree_thresh quantile of their module are considered hub TFs.
The degree of a TF is its intramodular connectivity, as in WGCNA: the sum of its network adjacencies with the
other TFs of the same module. The adjacencies are computed as in compute.WTCNA(), with the same
soft-thresholding power, network type and correlation (for a "signed" network, \(((1 + r)/2)^{power}\), so a
negative correlation adds almost nothing instead of being subtracted).
Arguments
- network
TF network obtained from
compute.WTCNA().- mm_thresh
Threshold for module membership. Default is 0.8.
- degree_thresh
Quantile threshold for degree within each module. Default is 0.9 (top 10%).
Value
A list with two elements:
- hubGenes
Named list of hub TFs per module. Modules without hub TFs are left out (a message lists them).
- detailedData
Data frame with one row per hub TF (TF names as row names) and the columns
Module,ModuleMembershipandDegree.
Details
The TF activities are taken from network (element TFs_matrix) and matched to the module scores
by sample name. For a multi-cohort network (compute.WTCNA() with batch), the per-cohort TF activities
are stacked to compute the membership, and the degree is the mean of the per-cohort degrees (each with the power
of its cohort).
Networks built before the soft-thresholding power was stored in network (element Network parameters)
have their power chosen again with the rule of compute.WTCNA() (with a message), assuming a signed network
and Pearson correlation.
Examples
data("network.tuto")
hub_tfs <- identify_hub_TFs(network.tuto, mm_thresh = 0.8, degree_thresh = 0.9)