lhe17
commented
1 year ago Hi Zhaohui,
Thank you for your questions.
"I suspect this issue arises from the difference of the input count: raw for NEBULA and log2 normalized for MAST."
Yes. Because these two tools are based on different models, it might not be appropriate to directly compare the logfc.
"What do you suggest I do in this regard? To be more specific, if I use the natural base logFC from NEBULA results, what is the appropriate cutoff I should apply?"
I'm sorry that I do not have a suggestion for this. I'm not aware of a universal cutoff for logFC. I think it might depend on how much change of the expression you expect to see in your experiment or with one unit increase in your variable. logFC=0.2 means that one unit increase of this variable will result in exp(0.2)=1.22-fold increase of the expression on average, which might be a lot in some experiments, but can be nothing special in others.
Best regards,
Liang
On 1/4/2023 5:26 PM, ZHAOHUI MAN wrote:
Hi Liang, I understand the logfc in the result is natural based. However, for single cell analysis, I usually use log2fc as a cutoff to select DEGs. When I use NEBULA with raw counts as input, I converted the result to log2fc and found the values are much bigger than those generated by other method like MAST. For MAST, I used log2 normalized counts as input and the cutoff for the resulting log2fc i used is normally 0.2. If I still use this cutoff for NEBULA after the logFC is converted to log2FC, it doesn't work for all the genes have log2fc greater than 0.2. I suspect this issue arises from the difference of the input count: raw for NEBULA and log2 normalized for MAST. What do you suggest I do in this regard? To be more specific, if I use the natural base logFC from NEBULA results, what is the appropriate cutoff I should apply? Thank you very much. Best. Zhaohui Man
— Reply to this email directly, view it on GitHub https://github.com/lhe17/nebula/issues/14, or unsubscribe https://github.com/notifications/unsubscribe-auth/AGDISUWXIY4X6CXDDC724RDWQWQBZANCNFSM6AAAAAATQ7H4YE. You are receiving this because you are subscribed to this thread.Message ID: @.***>
MANZHAOHUI
Hi Liang, I understand the logfc in the result is natural based. However, for single cell analysis, I usually use log2fc as a cutoff to select DEGs. When I use NEBULA with raw counts as input, I converted the result to log2fc and found the values are much bigger than those generated by other method like MAST. For MAST, I used log2 normalized counts as input and the cutoff for the resulting log2fc i used is normally 0.2. If I still use this cutoff for NEBULA after the logFC is converted to log2FC, it doesn't work for all the genes have log2fc greater than 0.2. I suspect this issue arises from the difference of the input count: raw for NEBULA and log2 normalized for MAST. What do you suggest I do in this regard? To be more specific, if I use the natural base logFC from NEBULA results, what is the appropriate cutoff I should apply? Thank you very much. Best. Zhaohui Man