For every single dataset from GDS or recount2 cohort and for a given \(k\), a dendrogram that measures the similarity between partitions from 24 methods (4 top-value methods by 6 partitioning methods) was first generated by cl_dissimilarity()
function from clue package. Then an ensemble of dendrograms from all \(N\) datasets from the cohort was constructed and a consensus dendrogram was generated by cl_consensus()
function from clue package. Figure S5.1 illustrates the consensus dendrogram for each k and for each cohort.