Closed AngCamp closed 1 year ago
AngCamp
commented
1 year ago Also is it then safe to just run multiple test correction on the p-values of the interaction of interest to determine DEGs? I have just been removing genes which nebula returns nan's in the pvalues, is that best?
lhe17
commented
1 year ago Hi AngCamp,
Yes, I think you can analyze all the cell types at once. If you have too many variables in dkkl1.nebula.df, it can be undersampled.
Best regards, Liang
On Fri, Jun 23, 2023 at 8:29 PM AngCamp @.***> wrote:
I am reanalyzing an experiment which had ~3k cells sequenced at an average depth of 1.5 million reads. There were 7 cell types of interest in several conditions. I am specifically interested in an interaction between the condition and a lable the cells express. So right now I am analyzing each cell type in separate objects. The model matrix is as follows:
dkkl1.nebula.df = model.matrix(~Condition:Label, data=dkkl1_nebula$pred)
Do you think I could do the following and analyze all the cell types at once?
dkkl1.nebula.df = model.matrix(~Condition:Label:celltype, data=dkkl1_nebula$pred)
Some of the replicates only had 6 of the cell types in the interaction of interest. Will nebula still be able to perform well with this data, or is it just too undersampled?
— Reply to this email directly, view it on GitHub https://github.com/lhe17/nebula/issues/26, or unsubscribe https://github.com/notifications/unsubscribe-auth/AGDISUSCTXYXGSN2BUFUPKDXMXOCDANCNFSM6AAAAAAZR4URB4 . You are receiving this because you are subscribed to this thread.Message ID: @.***>
lhe17
commented
1 year ago Hi AngCamp,
Yes, you can run multiple test correction on the p-values after removing genes with nan.
Best regards, Liang
On Fri, Jun 23, 2023 at 9:01 PM AngCamp @.***> wrote:
Also is it then safe to just run multiple test correction on the p-values of the interaction of interest to determine DEGs? I have just been removing genes which nebula returns nan's in the pvalues, is that best?
— Reply to this email directly, view it on GitHub https://github.com/lhe17/nebula/issues/26#issuecomment-1604735452, or unsubscribe https://github.com/notifications/unsubscribe-auth/AGDISUQWI46X2MMTNRF6PSLXMXRZBANCNFSM6AAAAAAZR4URB4 . You are receiving this because you are subscribed to this thread.Message ID: @.***>
I am reanalyzing an experiment which had ~3k cells sequenced at an average depth of 1.5 million reads. There were 7 cell types of interest in several conditions. I am specifically interested in an interaction between the condition and a lable the cells express. So right now I am analyzing each cell type in separate objects. The model matrix is as follows:
dkkl1.nebula.df = model.matrix(~Condition:Label, data=dkkl1_nebula$pred)Do you think I could do the following and analyze all the cell types at once?
dkkl1.nebula.df = model.matrix(~Condition:Label:celltype, data=dkkl1_nebula$pred)Some of the replicates only had 6 of the cell types in the interaction of interest. Will nebula still be able to perform well with this data, or is it just too undersampled?