Is it good to use differentially expressed genes and differentially methylated regions for DIABLO analysis?

I am not sure whether it is ok to use the differentially expressed gene and methylation sites for DIABLO. I saw 2 publications, they did differential expression and methylation analyses before data integration and only integrated the DE genes and DMRs instead of using all genes and methylation sites. Do you reckon that I should do the same way? Thank you so much!

hi @Casey

it depends on the overarching aim of your analysis. If you are interested in prediction, then inputting DE features will create an optimistic bias in the analysis. If you want to further explore associations between DE features then you could (in general this is not my preferred choice, as DIABLO is good to sometimes unravel strong associations between features that are not necessarily differentially expressed).


1 Like

Hi Kim-Anh, thank you for your reply; it has been very helpful!