First compiled: May 5, 2017.
See the notebook.
Scanpy provides a number of Seurat's features (Satija et al., Nat. Biotechnol., 2015), but at significantly higher computationally efficiency. Here, we reproduce most of Seurat's guided clustering tutorial as compiled on March 30, 2017. The tutorial starts with preprocessing and ends with the identification of cell types through marker genes of clusters. The data consists in 3k PBMCs from a Healthy Donor and is freely available from 10x (here from this webpage). The profiling information for Seurat has been obtained within seurat_R.ipynb.
Note: The profiling information was obtained in June 2017 for Scanpy 0.2.1 and Seurat 1.4.0.4. In the meantime, both Scanpy and Seurat have become faster and the difference should not be as dramatic any more.
| Scanpy | Seurat | |
|---|---|---|
| preprocessing | < 1 s | 14 s |
| highly variable genes |  |
 |
| correction, regressing out | 6 s | 129 s |
| PCA | < 1 s | 45 s |
 |
 |
|
| clustering | 1.3 s | 65 s |
| tSNE | 6.5 s | 25 s |
 |
 |
|
| finding marker genes | 0.8 s | 96 s |
 |
 |
|
 |
 |