small RNA differences between sequencing deeper or two times

Question

[amanda.fernandez.r]

Dear all, I´m new with small RNA sequencing data. I just have a problem, I have 96 sampes of three different conditions. They have been prepared with smallRNA Truseq library (Illumina) and sequenced with V3 chemistry in 4 different pools in 4 lanes (HISeq 2500) in order to get around 10 millions of reads from each sample. After sequencing, one of the pools (24 samples) yielded less reads than expected (sometimes less than a half than the others, around 5 M). These samples were resequenced but still, similar reads were obtained and core facility tell me to merge the reads with the previous ones. However, I´m not sure about this... My question is that since I´m looking also for new/rare miRNAs, wich are usually in low frequency, I cannot detect them with a 2 x 5 million reads, instead of 10 M in one run, right? Additionally, the 24 samples with this problem belongs to one condition in my experiment, therefore now I cannot compare to the other two conditions ? I really appreciate your comments, thanks!!

Answer 1

[WouterDeCoster]

since I´m looking also for new/rare miRNAs, wich are usually in low frequency, I cannot detect them with a 2 x 5 million reads, instead of 10 M in one run, right?

In both cases you are sampling reads from a library, and the outcome should more or less be the same. The chance for a read/gene to get sequenced is roughly proportional to its abundance. Read counts follow a Poisson distribution (sampling).

24 samples with this problem belongs to one condition

So the samples belonging to one condition were also pooled together and sequenced? That's a nice way of creating a batch effect, I'm afraid. This stratification should be avoided and all conditions should be mixed across pools/lanes/...

These samples were resequenced but still, similar reads were obtained

Was a new library prepared or the same pool sequenced again? Was the obtained library yield (DNA quantitiy) comparable to the other pools?

Answer 2

[amanda.fernandez.r]

Thank you very much for you answer. The 24 samples were pooled before size selection, therefore the same pool of libraries were re-sequenced, sadly not a new library... I think the problem was the quentification of the library, but I´m working with a core facility and I cannot control this process. I have tell them about the batch effect, but I´m afraid they did not mix the conditions within pools/lanes... Any suggestion to fix this problem with the analysis? I really appreciate your advice, thank you again!

[WouterDeCoster]

Please use ADD COMMENT to reply to earlier posts, as such this thread remains logically structured and easy to follow.

I would expect the issue to be technical, i.e. in the library preparation. As you suggest, the quantification could have been off, leading to adding less pooled library to the flow cell than should have been added, therefore sequencing less. Did you get information about the cluster density and the % of clusters passing filters for the lanes?

Were the 24 libraries also quantified before pooling? Was the concentration lower at that point?

Did the core facility give you the remainder of the libraries or pools after sequencing?