Skip to main content

Table 3 Summary of comparing Misty Mountain with state of the art flow cytometry specific clustering methods

From: Misty Mountain clustering: application to fast unsupervised flow cytometry gating

Data set Manually gated 2D barcoding& Simulated 5D Gaussians Simulated 2D non-convex 3D rituximab 4D GvHD Manually gated 4D
OP9
Misty Mountain accuracy sens (%) 100 100 100 - - 100
   spec (%) 100 100 100 - - 100
  CPU (sec)   10 196 6 0.3 0.8 3.6
FLAME accuracy sens (%) 20a 60b - 0d* 100d - - -
   spec (%) 33a 50b - 0d* 100d - - -
   CPU (sec) 5.104 >3.105 1.104 10 360 1.4 · 104
flowClust accuracy sens (%) 45a* 60b* 100c 0c* 100d - - 60d* 60*
   spec (%) 60a* 55b* 100c 0c* 100d - - 75d* 38*
  CPU (sec)   5.104 4.104 7200 43 480 3660
flowMerge accuracy sens (%) 25 100 0 - - 80
   spec (%) 45 100 0 - - 57
  CPU (sec)   1.3 · 105 1.27 · 105 7200 124 1020 8400
flowJo accuracy sens (%) 45 - - - - -
   spec (%) 47 - - - - -
  CPU (sec)   1-10 - - 1-10 1-10 -
  1. a optimal cluster number: 12
  2. b optimal cluster number: 24
  3. a*optimal cluster number: 15
  4. b*optimal cluster number: 22
  5. c optimal cluster number: 5
  6. c* optimal cluster number: 2
  7. d optimal cluster number: 1
  8. d* optimal cluster number: 4
  9. * optimal cluster number: 8
  10. &to save CPU time a data set, reduced by 80%, has been analyzed by FLAME, flowClust and flowJo
  11. sens (sensitivity) = (# of correctly assigned clusters)/(# of clusters in gold standard)
  12. spec(specificity) = (# of correctly assigned clusters)/(total # of assigned clusters)
  13. Gold standards were independent expert manual clustering for experimental data and specified clusters for simulated data.