The performance of different methods on the simulated datasets.

(A) The clustering runtime vs. the number of cells in the simulated datasets for all four methods. (B) The memory usage vs. the number of cells in the simulated datasets for all four methods. (C) The estimated number of clusters vs. the number of cells in the simulated datasets for three out of four methods: Secuer, Louvain and Leiden. k: thousand, M: million.