Question: I am not satisfied with number of cells that Cell Ranger identified. How can I redo the analysis using a specific number of cells to bypass the cell detection algorithm?
Answer: First of all, the cell calling algorithm in Cell Ranger works well in most cases. We recommend using force-cells option only if the number of cells estimated by Cell Ranger is not consistent with the barcode rank plot.
There are two ways to redo the analysis using a specific number of cells:
(1) Rerun cellranger count or cellranger multi with the force-cells option. Detailed documentation can be found here: count, multi.
(2) Alternatively, you can rerun the analysis with cellranger reanalyze pipeline using the --force-cells parameter. If you are specifying a larger number of cells than were originally called, --matrix must specify the raw gene-barcode matrix H5 file; if you are calling a smaller number, use the filtered gene-barcode matrix H5. The secondary analysis steps will be re-run using the updated set of filtered barcodes. Example:
cellranger reanalyze --id=my_reanalysis --force-cells=10000 \
--matrix=my_pipestance/outs/raw_gene_bc_matrices_h5.h5
Warning: although running reanalyze pipeline may save some time, the output web summary from reanalyze will not includes all metrics (such as mapping metrics) and plots (such as barcode rank plot), because the input is matrix, not raw data. If these metrics and plots are important for troubleshooting, please rerun with count or multi pipeline.