search

ebi-gene-expression-group / snRNA-mapping-rate

Comparison of Mapping rates for snRNA-Seq datasets with or without introns in reference and different mapping tools
0 stars 0 forks source link

Cool project! Some thoughts. #1

Closed rob-p closed 2 years ago

rob-p commented 2 years ago

Hi,

I randomly stumbled across this project and it seems pretty interesting! I just had some basic thoughts about how it looks like alevin-fry is currently being run. One note is that when producing a RAD file for consumption by alevin-fry (e.g. here the --keepCBFraction, --freqThreshold, and --tgMap parameters are unnecessary. They affect the behavior of alevin itself during quantification, but when alevin-fry is being used for quantification they have no effect, since alevin is essentially just doing mapping of the reads to the target transcriptome.

The other thought is that you may want to use the --unfiltered-pl flag during permit list generation, rather than the knee method here to make the results as comparable as possible to other methods. Specifically, the other tools (apart from the original alevin) make use of the external permit-list (i.e. the so-called whitelist) when processing 10x chromium data. The knee method performs filtering to derive the permit-list from the data itself, and works when there is no external list of potential barcodes available, but will produce a more stringent filtering (sometimes considerably) than quantifying with the --unfiltered-pl option. Finally, the --min-reads <threshold> option can also be passed to the generate-permit-list step if you want to alter the default cutoff for the number of reads that must be present for a cell to be quantified in the output (the default is 10 in the current and recent versions of alevin-fry).

Best, Rob

pinin4fjords commented 2 years ago

Thank you very much for that input @rob-p, we really appreciate it!

For context, @nadjano is an intern we have attempting to address deficiencies in our scRNA-seq pipelines when dealing with snRNAseq. We are attempting to decide whether to just use a different intron-containing reference, or new tools/ pipelines.

There is of course a case for updating our Alevin-driven scRNA pipeline for alevin-fry as well, but that's a question for another day.

nadjano commented 2 years ago

Thanks a lot @rob-p! Your comments are really helpful 👍 I will definitely include them.

rob-p commented 2 years ago

Thanks! I'll be happy to provide feedback should any other questions come up as well (e.g. parsing/interpreting the output of alev-fry's USA mode).

DongzeHE commented 2 years ago

Hello all,

I just fixed a severe bug in the make_splici_txome function of pyroe. Please make sure that you use the latest version (0.4.0) of pyroe to make the splici references. I apologize for the inconvenience.

Dongze

nadjano commented 2 years ago

Thanks for the notification @DongzeHE! I will change to the latest version.

Best, Nadja