replace reComBat with limma::removeBatchEffect

[sreichl]

Consider replacing reComBat with limma::removeBatchEffect
https://rdrr.io/bioc/limma/man/removeBatchEffect.html

pro

• limma works, is maintained and widely used/tested
• we reuse the same package as for other steps (e.g., VOOM norm) and downstream analyses
• already provides intuition how the downstream linear model, given the same design, "sees" the data
• Works also with only continous covariates:
  • https://support.bioconductor.org/p/9153034/#:~:text=removeBatchEffect%20does%20take%20continuous%20variables,variables%20in%20the%20design%20matrix.
  • https://rdrr.io/bioc/limma/src/R/removeBatchEffect.R

con

• supports only two factors (batch1 and batch2)

Considerations

• required design matrix is tricky, unsure if I implement a formula or just "tell me the columns you care about as list" (and add them together into a formula internally).
• could also lead to a change in filter by expression step to support formulas or multiple columns?

decision: replace reComBat w/ limma:removeBatchEffect