Hello everyone!

My question is related to RNA-seq analysis utilizing DESeq2.

Input: I try to analyze DEGs between normal HEK293 cells and HEK293 depleted by one protein. So, I have 3 biological replicates. (2 of replicates I obtained in one day, and the third one in a week - condition was the same). I analyzed my data using DESeq2.

The problem: When I produce the PCA plot (top 500 genes) I have the situation when the first two replicates have a good clasterisation and thirds ones are away and formed another pair. So, I tried to analalyze data separetely - 1,2replicates together and the 3d replicate was splited on 2 bam-files to create counts table and also used in analysis. Then I analyzed the overlapping DEGs between 1,2 replicated and 3d replicate - there was overlapping of about 50% of genes. I also analyzed the intersection of DEGs between these two subexperiment and general analysis (when I used all 3 replicated together) and there was more than 50% overlapping. I think that differences between replicates are due to small differences of cells confluency when cells were transfected and this differences are reflected in PCA plot. But on the other hand I have a good intersection of DEGs.

So, could you please give me an andvise what the better way to analyse data in my circumstances. Is it better to analyse all 3 replicates together or is it better to analyse separetely and try to find overlaped DEG sets. Or maybe you can suggest me another way to analyze my dataset.

Thank you in advance, Best wishes, Alexander

ComBat function from the sva package is what you need. Refer to section 7 of the manual (ComBat to adjust for known batches). Or you can also include batch in your design matrix, for example ~ celltype + batch.