Hub selection in WGCNA

Question

[grant.hovhannisyan]

In WGCNA package, there is a function chooseTopHubInEachModule that selects top hub genes from each module of the network. Which filters and cut-off does it use for hub selection?

Thanks

Answer 1

[Kevin Blighe]

There are different ways to calculate hub scores. WGCNA's implementation is to simply return the gene with the maximum adjacency / 'connectivity' per module. Just type chooseTopHubInEachModule on your command line:

chooseTopHubInEachModule
function (datExpr, colorh, omitColors = "grey", power = 2, type = "signed", 
    ...) 
{
    isIndex = FALSE
    modules = names(table(colorh))
    if (!is.na(omitColors)[1])) 
        modules = modules[!is.element(modules, omitColors)]
    if (is.null(colnames(datExpr))) {
        colnames(datExpr) = 1:dim(datExpr)[2]
        isIndex = TRUE
    }
    hubs = rep(NA, length(modules))
    names(hubs) = modules
    for (m in modules) {
        adj = adjacency(datExpr[, colorh == m], power = power, 
            type = type, ...)
        hub = which.max(rowSums(adj))
        hubs[m] = colnames(adj)[hub]
    }
    if (isIndex) {
        hubs = as.numeric(hubs)
        names(hubs) = modules
    }
    return(hubs)
}

The key lines are within the for loop.

Kevin

[grant.hovhannisyan]

Thanks, very helpful! So as far as I understand, chooseTopHubInEachModule selects only one hub per module? I got confused because of the name of the function, thinking that one hub per module is calculated by chooseOneHubInEachModule.

[Kevin Blighe]

I do not know what the utility is of the chooseOneHubInEachModule function. However, regarding the WGCNA package, there appears to be redundancy (in terms of code) across its various functions. Important to remember that WGCNA is not a Bioconductor package.