Dealing with two batches in RNA-seq

Question

[zizigolu]

Hi

I have RNA-seq from two batches; Lab technician says that he has run the RNA expression quantification two times in bathes 1 and 2. He says that one batch shows under counting and another one shows over counting. Now I have extracted two independent matrices of raw read counts from batches 1 and 2 but I don't know if I must run DESeq2 on each batch independently or I can take mean of read counts over two batches? Actually I am not sure what people do in this situation I just know he says everything has been the same in running experiments but he has run the RNA-seq two times

Thank you for any help

Answer 1

[Kevin Blighe]

Hey,

In your metadata, create a new variable called batch, and introduce this as a covariate in your DESeq2 design formula.

For example:

ID      batch   condition
sample1 first   case
sample2 first   control
sample3 second  case
sample4 first   control
sample5 second  case
sample6 second  control

Then, the design formula:

~ condition + batch

This will have the effect of adjusting the test statistics for batch.

Take a look at this very interesting thread: https://support.bioconductor.org/p/76099/#76101

Kevin

[zizigolu]

Sorry I have same samples for both batches 1 and batch 2 ; I mean experiment has been run two times on the same patients

For example tumor 1 in batch 1 and tumor 1 in batch 2

normal 2 in batch1 and normal 2 in batch 2

Does this make any difference in your suggested design?

[Kevin Blighe]

If they are from the same biopsy and library preparation, then just concatenate the FASTQ files together.

[zizigolu]

Yes from same biopsy of a patient (I am not sure about library preparation though) but I don't know why lab technician has run the experiment two times. He says one run shows over counting and one run shows under counting

[Kevin Blighe]

If it is tumour DNA, then 2 samples from the same biopsy can still be very different (tumour heterogeneity). Ultimately, processing them both separately and then combined will reveal more information to you.

[zizigolu]

Thank you Kevin, it is RNA-seq

I will do batch correction after processing them both separately then :)

[zizigolu]

@Kevin

This is my design for DESeq2

> head(mycols)
        condition batch
N_1_305         N     1
N_1_310         N     1
N_1_337         N     1
N_1_353         N     1
T_1_305         T     1
T_1_310         T     1
> tail(mycols)
        condition batch
T_2_337         T     2
T_2_338         T     2
T_2_344         T     2
T_2_346         T     2
T_2_349         T     2
T_2_353         T     2
>

By your package I got this bioplot by vsd values

![enter image description here][1]

From biplot, batch1 and batch2 of each patient are close to each other, Can I then collapse technical reads to each other?