Closed DrHogart closed 4 years ago
Thanks — I’ll address this tonight.
Thanks again for posting the issue, I think I've addressed this here:
c8d02c27c5af5f9a32951443c2f43b8e9fd2a068
In particular, I changed labelKeep
so it can be run on sparse matrices, and then added code to the vignette to clarify how and when to make the matrices dense.
https://github.com/mikelove/fishpond/blob/master/vignettes/swish.Rmd#L742-L748
swish
needs to compute ranks which currently makes the matrices not sparse, but the above should work for testing two groups of cells across a subset of expressed genes, which shouldn't be too large of a "cube" of data.
Thank you for the quick commit, it works on my data.
Out of curiosity, how large of data are you interested in testing, eg how many genes passing filtering (minimal expression) and how many cells at a time?
Also make sure to update tximport to >= 1.14.2.
We had a bug in importing inf reps but we pushed a fix.
My data actually is the sample- and UMI-barcoded bulk RNA-seq, so the number of samples (aka cells) is not huge, less than 50; there are 6000-8000 genes passing filters in each sample. My tximport is 1.14.2
Got it. Thanks. Shouldn’t be a problem converting to dense. It can get slow with ~1,000 of cells and all ~50,000 genes.
Oops i meant 1.14.2.
Hi, Currently the default output of
salmon alevin
is sparsed Matrix, but fishpond can deal only with regular matrix. E.g. trying to apply ofscaleInfReps
(v. 1.3.10) tosalmon alevin
(v.1.1.0) output file results to this: