ChromVAR on integrated dataset - suppressing batch effects ?

[solvi808]

Has anybody done batch correction in relation to ChromVAR analysis ?

I have an integrated dataset with samples of differing sequencing depth, etc. And I am seeing per sample enrichment of AP-1 , which I believe could be related to stress etc and I'm thinking the sequencing depth also puts a bias on the output.

I was wondering if there exists a recommended workflow for running ChromVAR in a way that the batch effect doesn't appear

So far my ideas have been : normalising per batch before ChromVAR , or batch correcting the output ChromVAR matrix itself , and finally another idea would be to regress/center out the sample per Motif .