Hello

According to the latest SVA vignette, the model matrix for the ComBat function should not include variables of interest, such as treatment/control but should include other covariates, such as sex. However, in previous vignettes and the SVA documentation the model matrix is defined as the "Model matrix for outcome of interest and other covariates besides batch". After trying both approaches using the vignette data, the q-values produced (again using the commands provided in the vignette) are quite different, with the model matrix consisting only of the intercept term is more conservative producing much larger values. It therefore seems that not including the variable of interest is the preferred approach. Would someone please clarify/confirm this?

Thank you

-Brett
 

Hi Brett,

It was pointed out to us that including the covariates in the correction step can lead to some anti-conservative bias in two-step procedures (where you first clean with ComBat, then do a separate differential expression analysis) so for now we are recommending not including the variables, but we are working on a more complete solution that will allow for the use of the variable of interest and correct the downstream analysis. Stay tuned!

Jeff