RNA-seq batches across different tissues

Question

[ATCG]

Hi,

I have RNA-seq data across different tissues, and I want to identify genes that are uniquely expressed in a tissue of interest. Some of these data were generated in the lab and sequenced in the same lane, and other data comes from public databases. Can I use DESeq2 normalization in this case? Is it appropriate to use sva in this case as done in:

http://127.0.0.1:30821/library/rnaseqGene/doc/rnaseqGene.html#removing-hidden-batch-effects

8.1 Using SVA with DESeq2

Thank you so much for your help!

[ivivek_ngs]

Do you know all the factors are causing the batch effect e.g. lane and other lab data? If your PCA is already showing variability due to lane effects and due to other labs, then you can perform the same using the COMBAT or edgeR removeBatcheffct() function. SVA is amazing and great when you are unable to know the technical sources of variation from your PCA/MDS plots using your metadata. It is then you perform the surrogate variable analysis to obtain a number of variables and then use it accordingly in the model.matrix for adjustment. However, be careful with the number of variables being used for adjustment in order to not overfit.