Is there a way to remove batch effect such that only the data from all batches are adjusted relative to one single batch, which remains unchanged? I have a primary batch and I want all my other batches to be corrected against that. But I don't want to impose a change in that primary batch data.
I've developed a new method, ConDo, which does this. You pass in a source batch which is transformed, and a target batch which is unaffected.
The downside is that you'll have to create a new CondoAdapter object to learn the transformation for each of the different source batches. But it shouldn't require a lot of code to do this. (Also, it's implemented in Python, not R.)
Preprint is here: https://arxiv.org/abs/2203.12720
Software is here: https://github.com/calvinmccarter/condo-adapter
Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
version 2.3.6
If I may ask, is there a particular reason why you want to do this? And in what context will you use the data corrected that way?
We have a reference cohort of genome wide methylation data that we have validated as having no change related to constitutional disorders. We will compare future batches to this primary cohort. We want to assume that the levels observed in this cohort are normal and most humans are like that, at least for certain genes that we are interested. We don't want to have these levels changed. I m wondering if there is a way to do this. Most methods for batch correction change the values from all batches.