I have a count matrix that underwent normalization to remove batch effects and technical variation. If I used DeSeq(after making values integers), would this be causing major errors?

I know from other posts that you should feed raw counts into Deseq, but I had issues with removing batch effects and technical variables that made it necessary to normalize the count matrix in order to clean up the data (among other steps).

If I now use Deseq, will that work, or do I need to write the code for each step by hand? Any other advice is much appreciated.

Best,

Hugo

It's better as a general rule to remove excess variability due to known factors (batch effects, etc) and unknown factors (various technical effects) as part of the linear regression, which will correctly account for the loss of degrees of freedom from having done so. You can fit batch as a fixed effect, and you can estimate changes due to unobserved technical effects using either the sva or RUV packages.

It depends. If you used a strategy like ComBat-Seq that preserve the integer nature and distribution of the raw counts then you could use DESeq2. If not then you probably completely altered magnitude of counts and distribution, hence the assumptions of DESeq2 are violated and results could be nonsense.