Cellranger alignment options / parameters

Question

[PianoEntropy]

I'm trying to align FASTA files with a custom genome reference (GTF file), which contains a few sequences for genes that were inserted. So far I've just run cellranger count on these files and everything exited without errors. However, in the downstream analysis the result is not as expected, and I am wondering whether this could have to do with the alignment. Basically, I am detecting the inserted sequences in many cells that should not contain it. One possibility is of course experimental error or contamination, but I was thinking it could also have to do with the alignment misaligning reads for those inserted sequences and detecting counts where there should not be counts.

Hence I was wondering if there are parameters to tune the "sensitivity" of the genome alignment in cellranger. Basically, I would want to be strict and discard reads which have even a small number of mismatches. Or if you have other suggestions to filter out potential misaligned reads, I'd be happy to know as well.

Cellranger count itself does not seem to have any relevant options, but it seems that on the background it runs STAR. Hence if it is possible to pass STAR arguments to cellranger, that could also be a possible solution (although I haven't found the relevant STAR parameters yet). Or maybe I'm just misunderstanding genome alignment altogether and there is not much to tune with the current algorithms?

Answer 1

[jared.andrews07]

There is no way to do this via cellranger without altering the source code. You could always trying other gene quantification/alignment options though, like alevin or STARsolo which would allow you full control over the parameters used. STARsolo even has directions for how to largely replicate cellranger results, so you could do that and tweak only your parameters of interest.