Replicates in RNA-Seq experiment

[keysoon]

Dear all,

This question is somehow linked to my previous thread http://seqanswers.com/forums/showthread.php?t=49153 . I am analyzing RNA-Seq data from three breed groups (FIN, TEX, FXT) with two diet conditions (FLU, CON). The experimental design is attached below(expdesign.png). Two additional libraries were prepared for one sample each from three groups (FIN-FLU, TEX-FLU, FXT-FLU), and one library from FXT-CON group was sequenced three times. This is further explained by the second figure, where number of reads sequenced as well as percentage of mapped reads are different in former case but almost same in later case (ReadsMapped.png). Can I treat samples for first three groups as technical replicates? Or, given the fact that both the number of reads and the amount of reads mapped are almost same, can I just choose one sample from (37C5, 37C6, 37C7) and treat all samples in former case as just biological replicates? I would be grateful to get your feedback and suggestion. I am using DESeq2 for the differential expression analysis.

Thank you.
Keysoon

[dpryan]

The 3 FXT-CON runs from the same sample can just be merged, they're technical replicates (or just use one of them, it shouldn't make a huge difference). Exactly which 3 groups are you talking about in terms of technical replicates? If they're different groups, then they're not technical replicates by definition. They're also generally not biological replicates, since those are typically within a single group (otherwise, what exactly are you replicating?).

[keysoon]

Thanks for your reply. 39A, 39E, 39J from FIN-FLU; 787, 4H, 4P from TEX-FLU and 7G, 7M, 4563 from FXT-FLU were prepared from one biopsy each. For example, 39A, 39E and 39J represent same biopsy but samples were prepared separately unlike 3 FXT-CON runs where single sample was sequenced three times.

[dpryan]

If they're from the same sample, they're technical replicates, it doesn't matter if they're from different libraries or the same one.

[keysoon]

Thank you. I was dubious because of the variation within replicates. 37C5, 37C6 and 37C7 behaved as if they are photocopies whereas other replicates looked as if they are different samples both in previous figure where I plotted total number of reads and % of reads mapped to ref. genome as well as in MDS-plot attached in this post. In the plot the replicates are 39A, 39E, 39J; 7G,7M,4563; 787,4H,4P; 37C5,37C6,37C7. I think the variation is mostly because the tissue is heterogenous.

[sarvidsson]

If you have variation due to sampling (e.g. heterogeneous tissue), which might be interesting to specify it as "biological replicates" (the term is very ambiguous, as the definition strongly depends on the context within which the data is analyzed) - this could help you to partition sampling vs treatment/time/whatever variable errors and study them.

Typically you would pool any data which is of purely technical nature (e.g. the same library sequenced in different lanes, two libraries made from the same RNA) as the variation (these days) is minor and not interesting to study.