search

willtownes / quminorm

Turning the quasi-UMI method into a bioconductor package
GNU Lesser General Public License v3.0
10 stars 2 forks source link

Can I use quminorm to normalize count table from 10x? #10

Open WubingZhang opened 2 years ago

WubingZhang commented 2 years ago

The total counts of single cells differ significantly, which affects the downstream differential expression analysis. So I was thinking to normalize the data before differential expression analysis. Seurat SCTransform partially solves my problem, and I think quminorm might be worth a try. Any suggestions?

Thanks!

willtownes commented 2 years ago

Thanks for your interest in our method! I would not recommend using quminorm on UMI counts. It is meant for read counts from protocols like Smart-seq2. It is not meant to fully normalize the counts, nor to make each cell have the same total count. So the pipeline we recommend would be apply quminorm to read counts, then apply whatever normalization you like to the quasi UMI counts, such as sctransform or scry. You could also just apply to the quasi-UMI counts a count-based dimension reduction like GLM-PCA, NewWave, scvi, or a newer method like countland.