search

z0on / GO_MWU

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

Proportion of "good" genes 0, but significant? #23

Closed magnew1 closed 3 months ago

magnew1 commented 3 months ago

Hello! I am running the GO_MWU analysis on oyster transcriptome responses to a virus using log2 fold change values. When I use the log2fold change values, some results give me significant GO terms but their proportion of "good" genes is 0, so I am confused how these can be significant values? I've attached an image of an example output three that I receive. Here's the code I input to get these results:

`gomwuStats(input, goDatabase, goAnnotations, goDivision, perlPath="perl", largest=0.1,
smallest=5,
clusterCutHeight=0.25

results=gomwuPlot(input,goAnnotations,goDivision, absValue=1, # un-remark this if you are using log2-fold changes level1=0.01, level2=0.001, level3=0.0001, txtsize=1.2,
treeHeight=0.5`

I set the levels to very small p-values because otherwise I ususally have >200 categories which makes the plot difficult to view. Is there a setting I'm missing?

Thank you! fam26_BP_lfc_plot

z0on commented 3 months ago

Hi Tori - this is not terribly surprising: this 0 just means that none of the genes in this GO term exceed 2-fold change magnitude, but this does not invalidate the MWU test results for the term. For example a "red" GO term can be significantly enriched with genes with positive log-fold changes even though none of them exceed 1 (absValue corresponding to 2-fold change). Misha

On Tue, May 14, 2024 at 12:45 PM Tori Agnew, PhD @.***> wrote:

Hello! I am running the GO_MWU analysis on oyster transcriptome responses to a virus using log2 fold change values. When I use the log2fold change values, some results give me significant GO terms but their proportion of "good" genes is 0, so I am confused how these can be significant values? I've attached an image of an example output three that I receive. Here's the code I input to get these results:

`gomwuStats(input, goDatabase, goAnnotations, goDivision, perlPath="perl", largest=0.1, smallest=5, clusterCutHeight=0.25

results=gomwuPlot(input,goAnnotations,goDivision, absValue=1, # un-remark this if you are using log2-fold changes level1=0.01, level2=0.001, level3=0.0001, txtsize=1.2, treeHeight=0.5`

I set the levels to very small p-values because otherwise I ususally have

200 categories which makes the plot difficult to view.

Thank you! fam26_BP_lfc_plot.png (view on web) https://github.com/z0on/GO_MWU/assets/64383219/c791d9d3-27aa-468a-8c90-2fd663fce5c8

— Reply to this email directly, view it on GitHub https://github.com/z0on/GO_MWU/issues/23, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABZUHGCCPZ2NNZMX4SYTFM3ZCJETPAVCNFSM6AAAAABHWUK7EWVHI2DSMVQWIX3LMV43ASLTON2WKOZSGI4TMMBTGM4TSMY . You are receiving this because you are subscribed to this thread.Message ID: @.***>

magnew1 commented 3 months ago

Great, thank you for clarifying!