Fig. 1

An overview of our pipeline for cluster reconciliation. Gene expression – or other equivalently structured molecular data – is input as a genes by samples matrix. This data is then fed through multiple consensus clustering algorithms (in this case, HC, KM and NMF) to produce multiple clustering solutions. The quality/robustness of the clusters is assessed. These clusters from multiple clustering methods are then reconciled to create “subtype communities” of similar clusters from across the algorithms’ solutions, by applying community detection to networks representing the similarity between clusters from all the algorithms