For downstream analysis from cellranger 10x scRNAseq eg. Seurat package, do you recommend using raw or filtered matrix?

Question

[simplitia]

Hi, its interesting that I have not seen this question come up yet. In cellranger 10x there are two folder, "raw_feature_bc_matrix" or "filtered_feature_bc_matrix". For downstream analysis such as those with the seurat package do you use the raw or filtered matrix and what is the rationale behind it? I'm inclined to use the raw since then I would have full control of the filtering but I want to know what others think.

thank you.

[rpolicastro]

Unless you have reason to think that droplet filtering failed, I usually load the filtered count matrix in. In most of my samples there's a clear inflection point from droplets with cells to droplets without, so I don't feel i would gain anything from manually filtering.

[simplitia]

rpolicastro thanks. After looking through several Seurat vignette's this is also at least what they used so I'm going to just stick with the filtered as well.