Is it necessary to adjust for multi-testing burden?

[Patricklv]

It is stated in the paper that increasing alpha (e.g., from 0.05 to 0.1) allows for identification of smaller modules at the risk of over-fitting. When using this algorithm to identify hierarchical modules, is it necessary to adjust for multi-testing burden? I want to maintain an overall type I error rate of 0.05, but it seems that the alpha in 'hierarchicalConsensus.m' is with respect to each node pair. But there are so many node pairs in a network. Not sure if it is necessary to adjust for this multi-testing burden and if yes, how should one determine the threshold? For reference, in genome-wide association study, the alpha level is set at 5e-6.