Question about model

[tbrunetti]

Hello,

First of all, thank you for publishing a paper on this method because I have felt that the pvalues of scRNA-seq just do not sit well with me and I have felt it produces a lot of false discoveries.

I just brought up this method/your paper to the attention of our informatics group, who are very interested in testing it, but one of the senior informatician had brought up a very valid point in that many of the scRNA-seq sets we deal with do not have normal distributions within each cluster and we may see bimodal or exponential distributions in our datasets. This is especially the case in our immunology data sets and dealing with rare cells.

1. Have you tested your new method on these types of data sets, since they would violate the normality assumption? Even Seurat's Wilcox test is non-parametric so not sure how the truncated normal fits in with non-parametric sets?
2. If not, is there an easy way to adapt your current use of MLE to feed into an exponential distribution or any non-parametric distribution for that matter instead of the truncated normal assumptions?

Any ideas or suggestions would be appreciated! Thank you!