search

z0on / GO_MWU

Rank-based Gene Ontology analysis of gene expression data
37 stars 17 forks source link

Question about Table of GO annotations and measures #18

Open Ruiqi-CUB opened 1 year ago

Ruiqi-CUB commented 1 year ago

Hi Misha,

I understand that you said It is important to have the latter two tables representing the whole genome (or transcriptome) - at least the portion that was measured.

I don't have a genome for my study system. I assembled a "meta-transcriptome" with all the tissues I sequenced for DEG quantification.

If I am going to do GO MWU in one tissue (for example, mantles), should I use the whole meta-transcriptome, or would it be more proper to use genes found expressed in the mantles?

Thanks a lot! Ruiqi

z0on commented 1 year ago

Hi Ruiqi - I meant all the genes that you subjected to statistical testing for differential expression, irrespective of whether they turned out to be significant or not. No need to include genes that were removed from the dataset because they were just not expressed at all. Does it make sense?.. Misha

On Mar 2, 2023, at 10:44 AM, Ruiqi-CUB @.***> wrote:

Hi Misha,

I understand that you said It is important to have the latter two tables representing the whole genome (or transcriptome) - at least the portion that was measured.

I don't have a genome for my study system. I assembled a "meta-transcriptome" with all the tissues I sequenced for DEG quantification.

If I am going to do GO MWU in one tissue (for example, mantles), should I use the whole meta-transcriptome, or would it be more proper to use genes found expressed in the mantles?

Thanks a lot! Ruiqi

— Reply to this email directly, view it on GitHub https://github.com/z0on/GO_MWU/issues/18, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABZUHGFAQWGUJK6ELRSWWLDW2DE7FANCNFSM6AAAAAAVNTWCFE. You are receiving this because you are subscribed to this thread.

Ruiqi-CUB commented 1 year ago

Thanks! Yeah that makes sense. My concerin is that even I used the genes not expressed in mantles for DE analysis, the results of pvalue will be NA if all the expression value is 0. Techinically they were not being tested, that's why I was thinking about remove those genes not expressed from the background. Does it sound reasonable?

z0on commented 1 year ago

We typically toss genes with mean count less than 3 across all samples and run DESeq on the rest. Please use log-fold changes for GO_MWU, not transformed pvalues as we did originally - turns out it creates bias toward highly expressed genes)

On Mar 2, 2023, at 11:10 AM, Ruiqi-CUB @.***> wrote:

Thanks! Yeah that makes sense. My concerin is that even I used the genes not expressed in mantles for DE analysis, the results of pvalue will be NA if all the expression value is 0. Techinically they were not being tested, that's why I was thinking about remove those genes not expressed from the background. Does it sound reasonable?

— Reply to this email directly, view it on GitHub https://github.com/z0on/GO_MWU/issues/18#issuecomment-1452223008, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABZUHGFTKXRUJQY73E6IZHDW2DIA5ANCNFSM6AAAAAAVNTWCFE. You are receiving this because you commented.

Ruiqi-CUB commented 1 year ago

I did use logFC for GO_MWU. I am saying use pvalue=NA to filter out the genes that present or not performed DE test on.