Table 1.
Performance (ARI and AMI) of different clustering methods from ten times analyses of the public human PBMC real dataset
| ARI | AMI | |
|---|---|---|
| Mean (SD), [range] | Mean (SD), [range] | |
| Single source clustering - RNA only | ||
| K-means | 0.524 (0.054), [0.508, 0.562] | 0.592 (0.022), [0.579, 0.632] |
| TSCAN | 0.563 (N/A), [N/A] | 0.591 (N/A), [N/A] |
| SC3 | 0.596 (0.082), [0.456, 0.709] | 0.629 (0.060), [0.513, 0.706] |
| DIMM-SC | 0.629 (0.030), [0.590, 0.676] | 0.666 (0.018), [0.653, 0.700] |
| Seurat | 0.564 (N/A), [N/A] | 0.660 (N/A), [N/A] |
| Single source clustering - protein only | ||
| K-means | 0.653 (0.093), [0.436, 0.760] | 0.732 (0.048), [0.632, 0.808] |
| TSCAN | 0.691 (N/A), [N/A] | 0.717 (N/A), [N/A] |
| SC3 | 0.649 (0.045), [0.568, 0.692] | 0.722 (0.035), [0.670, 0.777] |
| DIMM-SC | 0.698 (0.046), [0.619, 0.741] | 0.744 (0.025) [0.701, 0.766] |
| Seurat | 0.726 (N/A), [N/A] | 0.757 (N/A), [N/A] |
| Multi-source joint clustering | ||
| K-means | 0.661 (0.107), [0.471, 0.811] | 0.723 (0.059), [0.577, 0.781] |
| TSCAN | 0.586 (N/A), [N/A] | 0.635 (N/A), [N/A] |
| SC3 | 0.679 (0.162), [0.430, 0.911] | 0.723 (0.096), [0.595, 0.870] |
| MV-NMF | 0.701 (0.154), [0.380, 0.856] | 0.679 (0.098), [0.473, 0.779] |
| PC-MSC | 0.641 (0.029), [0.621, 0.723] | 0.679 (0.018), [0.672, 0.729] |
| BREM-SC | 0.728 (0.091), [0.585, 0.840] | 0.737 (0.048), [0.674, 0.800] |