7.3 Run MICA

MICA features two different modes named by their different dimension reduction techniques:

• Multi-Dimensional Scaling (MDS) mode: this mode is more accurate and robust for small datasets (less than 5,000 cells, be default) due to its global dimension reduction nature;
• Graph Embedding (GE) mode: this mode works better with large datasets (more than 5,000 cells, by default) using a graph embedding approach to explore distant neighbor cells.

In this case, since there are 13,605 cells, we will use the MICA GE mode for the clustering:

mica ge -i /work-path/PBMC14k/MICA/micaInput.txt -o /work-path/PBMC14k/MICA/micaOutput -minr 0.1, -maxr 9.0 -ss 0.05 -nw 40

This command will generate the clustering results of multiple resolutions, from 0.1 to 9.0, with a step size of 0.05.