My understanding of the challenge scoring wiki page is that the submitted modules will be treated as an unordered set. That is, unlike other DREAM challenges where participants submit ranked lists that are evaluated with area under the precision-recall and ROC curves, here the module identifier and confidence score will not be used to rank predictions for the primary score. The confidence score may be used for secondary analyses, but even then the module identifier is not considered for tie-breaking. Is that correct? Also, when computing false discovery rate is the total number of hypothesis tests (number of modules predicted) or (number of modules) * (the number of GWAS gene datasets)? I would expect the latter.