Hello, I am working with RNAseq data which was acquired from experiments as following:

• PBMCs were extracted from blood of healthy donors, then extracting monocytes, pooling monocytes of different healthy donors, then differentiate pooled monocytes into macrophages.
• After that, Mycobacterium tuberculosis isolates were infected into macrophages, I also included lab strain H37Rv, and no-infection group.
• Then macrophages were collected RNA at 4h and 24h post-infection.
• Because the heavy workload, I divided into 2 batches, each batch included 2 samples of no-infection group, 2 samples of H37Rv as control.
• Macrophages from 2 batches still from the same heathy donors.
• After collecting all RNA samples, RNA was sequenced at the same time.

When I did PCA, there was big difference between batches, even for no-infection and H37Rv group as picture below

I would like to ask that there are any method for batch effect removal or normalization to combine the data between batches.

I was thinking to normalize Mtb strain to no-infection group in each batch before comparing between isolates.

Thank you in advance.

Since batch is orthogonal to the treatment, you just add a batch factor to your model.

ComBat is appropriate for this type of design and BatchQC can be used to evaluate how well ComBat corrects for the batch effect. BatchQC provides a shiny interface where you can normalize, batch correct and view multiple visualization to determine the effectiveness of batch effect correction. Your batch-corrected (and/or normalized) data can then be exported for downstream use.