ComBat correction with one batch containing only one sample - keep or remove?

Question

[Sebastian Hesse]

I am using sea analysis and combat to correct a proteome dataset. During the correction process ComBat tells me that one batch contains only one sample. Apparently this leads to a modification of the method used:

Note: one batch has only one sample, setting mean.only=TRUE

My question is of this is ok or if I should rather remove this one sample from my analysis so ComBat can run normally?

Thanks a lot! Sebastian

[miaowzai]

I use ComBat for batch effect removal, too. From my understanding, it removes the batch effect through adjusting the data mean and variance from different bathes. If one of your batches contains only one sample, there is no way that ComBat can remove the variance (since there is no variance), and it has to set mean.only=T. If that one-sample batch is important to your analysis, in my opinion, it is totally ok to keep it and use the mean.only=T ComBat. If you have a lot of samples and you do not care about throwing away that one sample, then removing it you should be able to use mean.only=F setting. I would suggest doing PCA before and after you apply ComBat and see how ComBat has changed your data. That might give you some idea of whether including or excluding that one sample.

[Sebastian Hesse]

Thanks, I removed it form the analysis at the end.