Adjustment by several variables, sequential adjustment or combine them in a single one?

Question

[ev97]

My RNA data has to be adjusted by 3 variables. I have seen different methods to use:

• ComBat-Seq (sva package)
• limma (removeBatchEffect function)

On the one hand, limma allows you to adjust by several variables.

On the other hand, ComBat-Seq only allows you to adjust by one single batch, therefore, if you want to adjust by more than 1:

• You run sequential adjustments.
• Combine the variables (e.g. 3) in one single variable.

The first method requieres log2 normalized counts... and the second handles raw counts.

However, I don't know what it is the most accurate (and better) procedure to adjust by batch. Should I adjust by the 3 variables in one go, combine the 3 variables in one.... or run sequential adjustments?

Question: From your point of view/experience, what do you think or how do you usually do these type of procedure?

I would really appreciate any feedback.

Thanks very much in advance