Skip to main content

Table 2 Selection of the optimal level in the nSBM hierarchy

From: Nested Stochastic Block Models applied to the analysis of single cell data

Dataset Cells D \({\hat{k}}\) \(i^k\) \(B_k\) \(i^Q\) \(B_Q\)
sc-mixology [47] 860 5 21 1 6 1 6
Chromium 10x [54] 1523 8 43 0 58 1 13
Quartz-seq2 [54] 1266 8 37 0 62 1 12
MARS-seq [54] 1401 9 9 1 16 1 16
iCELL8 [54] 1830 9 20 1 21 2 6
Mouse brain [50] 2688 15 8 2 8 1 23
Planaria [10] 21,612 \(51^*\) 34 2 22 3 10
  1. For each dataset we report the number of groups D that were given by the authors. The optimal level selection should recover a number of groups in the order of magnitude of D. Value of D in Planaria dataset is derived from manual curation of Louvain clustering. \({\hat{k}}\): number of groups according to RMT, \(i^k\): level selected according to RMT criterion, \(B_k\): number of partitions at level \(i^k\), \(i^Q\): level at which modularity is maximal, \(B_Q\): number of groups at level \(i^Q\)