search

SONGDONGYUAN1994 / ClusterDE

A post-clustering differential expression (DE) method robust to false-positive inflation caused by double dipping
https://songdongyuan1994.github.io/ClusterDE/
Other
23 stars 4 forks source link

constructNull function generate counts with Nan values #2

Open PeterZZQ opened 11 months ago

PeterZZQ commented 11 months ago

Hi,

I'm using the fast version of constructNull function, and the generated data have "NA" in it. The input matrix is a UMI count matrix, the command that I used is ClusterDE::constructNull(counts.cluster.age, nCores = 8, fastVersion = TRUE). May I know how to fix it? Should I make all "NA" to be 0s?

SONGDONGYUAN1994 commented 11 months ago

Hi Ziqi, Can you describe the NA in more detail? How many genes have NA, and the NAs are for all cells or just a subset of cells? Usually, the NA comes from the terrible fit of NB distribution. You can change it to all 0, but this is definitely a compromise. If you can let me know when it happens, I can fix it.

Best, Dongyuan

PeterZZQ commented 11 months ago

Hi, Yes, the total number of NA within the synethetic null:

image

And for every cell, there is always the same 175 genes with NA values

image

If I set the fastVersion to be FALSE, then there is no NA in the synthetic null.

SONGDONGYUAN1994 commented 11 months ago

Hi Ziqi, Could you share a demo data? I hesitate that these genes are almost all zeros, but I need to check it further.

Best, Dongyuan

PeterZZQ commented 11 months ago

Hi Dongyuan,

Currently it is hard to shared the data. I double checked the counts of these genes, these genes indeed have most of their counts to be 0. This could be the reason. I will stick to the fastVersion=FALSE, it seems to work well.