[DESeq] Collapsing replicate across sequencing batches?

Question

[Kristin Muench]

Hello,

My colleague performed a 3' RNA-Seq experiment twice. She ran a subset of these samples twice. Can I collapse these like technical replicates using the DESeq collapseReplicates()? If not, should I remove them from my study because 'doubling up' on samples will bias the results?

When I say she ran a subset of these samples twice, I mean she had a vial of RNA for those samples, and went through the whole library preparation/sequencing steps about xis months apart. This doesn't seem like a true technical replicate because the library prep as well as the sequencing were done on a different date. However, it's not exactly a biological replicate, since it is the same sampled sequenced twice, not another unique organismal sample in that condition.

Thanks for any advice you have on how we should handle this situation!

[h.mon]

If the technical replicates cover all treatments, then they can be included as a batch effect in the model.

[Kristin Muench]

They do cover all treatments, but not all samples. Would the batch variable look something like:

SAMPLE         | EXPERIMENT DAY               |  BATCH
------------   | ---------------------------  | -----------
A              |  Day 1                       | Original
A              |  Day 2                       | Rep
B              |  Day 1                       | Original
C              |  Day 1                       | Original
C              |  Day 2                       | Rep
D              |  Day 2                       | Original

I didn't think I could use "Sequencing Day" as the batch variable because there are some samples (e.g. D, above) that only appear on the second day and don't have a replicate.

Also, are they still technical replicates if they were created during a different library preparation?

Thanks for your help.