Skip to main content

Table 2 Mean performance on the ten datasets of each method in terms of AMI and silhouette

From: Tuning parameters of dimensionality reduction methods for single-cell RNA-seq analysis

Method Mean AMI Mean silhouette
  Default Best ANOVA AMI heuristic Silhouette heuristic Default Best ANOVA AMI heuristic Silhouette heuristic
scran 0.840 0.868 0.841 0.741 0.362 0.547 0.396 0.494
Seurat 0.788 0.860 0.814 0.683 0.369 0.543 0.490 0.373
ZinbWave 0.780 0.896 0.851 0.825 0.249 0.609 0.562 0.591
DCA 0.758 0.885 0.837 0.583 0.396 0.639 0.403 0.381
scVI 0.560 0.872 0.709 0.510 0.384 0.621 0.482 0.318
  1. The “Default” columns correspond to the performance of each method using its default parameters, with a dimension of 10. The “Best” column corresponds to the best performance reached after varying the parameters. The “ANOVA AMI heuristic” column corresponds to the performances of the new default parameters described in the “Influence of parameters on performance” section. The “Silhouette heuristic” column corresponds to the performance of the heuristic described in “Tuning parameters in practice” section