Table 2 Average NMI score from different clustering methods on the datasets (standard deviation of scores in parentheses) along with optimal resolution parameter or choice of k used in each method
Dataset | K-means on PCA space | Leiden | Louvain | K-means on Cellograph-derived latent space |
|---|---|---|---|---|
Organoid | \(k=8\) 0.124 (0.019) | Resolution = 0.2 0.183 (0.071) | Resolution = 0.4 0.183 (0.021) | \(k=2\) 0.513 (6.695 \(\times 10^{-16}\)) |
Drug holiday | \(k=4\) 0.706 (9.255\(\times 10^{-4}\)) | Resolution = 0.2 0.817 (0.018) | Resolution = 0.2 0.824 (3.956\(\times 10^{-3}\)) | \(k=5\) 0.805 (1.292\(\times 10^{-4}\)) |
Myogenesis | \(k=10\) 0.538 (0.013) | Resolution = 0.1 0.715 (0.016) | Resolution = 0.5 0.655 (0.018) | \(k=7\) 0.87 (5.174\(\times 10^{-5}\)) |