Help and SupportPlease consult the
scdeGoogleGroup forum for common questions. Join the group to post new questions. To submit a bug report or enhancement request, please use the Github issue tracker.
Q: Can I use SCDE/PAGODA with UMI data?
A: Yes. In the error building step, you will need to reduce
min.count.threshold to 1.
Q: Can I use SCDE/PAGODA with non-count data?
A: SCDE requires count data for its error modeling. If you would like to use PAGODA with normalized non-count data, see this experimental tutorial for more information.
Q: Can I use SCDE/PAGODA with 1000s of cells?
A: Yes, however we have set default options based on a 96 cell sample. Therefore, parameters in the error modeling step such as
min.nonfailed will likely need to be increased to accommodate.
Q: How can I integrate my ERCC spike-ins for normalization?
A: You can use your ERCC spike-ins to fit the dependency between expression magnitude and variance:
# perform varnorm based on ercc fit ercc.cd <- cd[ercc.genes,] ercc.varinfo <- pagoda.varnorm(knn,counts=ercc.cd,trim=3,ncol(ercc.cd),fit.genes=ercc.genes,n.cores=n.cores,plot=T,verbose=1,max.adj.var=5)
The resulting plot will show the spike-ins in green. However, n our experience, the spike-ins often show systematic deviations from the genome wide trend, showing either significantly higher or lower variance. The former, in particular, suggests that they may not be representative of the technical noise impacting the real mRNAs so we recommend proceeding with caution.