Figure 2
From: scCAN: single-cell clustering using autoencoder and network fusion
Absolute log-modulus values obtained from CIDR, SEURAT3, Monocle3, SHARP, SCANPY, and scCAN for 27 real scRNA-seq datasets. This metric measures the difference between the number of clusters and the number of true cell types. The average log modulus of scCAN is 0.59 while those of Monocle3, SCANPY, SHARP, SEURAT3, and CIDR are 1.35, 1, 0.72, 0.64, and 0.63, respectively. scCAN significantly outperforms other methods by having the smallest absolute log-modulus values (Wilcoxon p-value of \(p=8.6\times 10^{-4}\)). Note that the dataset Brain 1.3M was excluded from this analysis because it does not have true cell type information.
