After initial manual assessment, we excluded some of the clusters from further consideration based on their small size in terms of number of genes or gene sets. We left out clusters that had fewer than 150 genes total when we merged the genes in all their member gene sets to allow for sufficient number of genes for subsequent refinement by meta-analysis. We also removed clusters with fewer than six gene sets as the smaller clusters usually lacked sufficient information (in terms of descriptions of and overlaps among their constituent gene sets) to deduce meaningful biological theme. The filtering left 168 clusters for the next stage.
