Adding RNA-Seq reads from one sequencing experiment to reads from another experiment

Question

[newbio17]

We outsourced some RNA-Seq experiments to a 3rd-party vendor. When we first received our data, they notified us that the number of reads generated did not match the number of reads guaranteed in the quote. So they told us that they will do additional sequencing if we agreed to wait 5-10 days. The final data we received consisted of reads coming from 2 RNA-Seq experiments combined into 1 pair of reads (R1.fq, R2.fq) instead of 2 pairs of reads (A_R1.fq, A_R2.fq, B_R1.fq, B_R2.fq) with only ~5 million coming from the second experiment.

Q: Is it okay to add reads from one RNA-Seq experiment to reads from another RNA-Seq experiment if they were done on the same sequencing platform and used the same libraries?

The place where we outsourced our experiments says that we should be okay for the following reasons:

• They carry out experimental technology self-assessment regularly (not too sure exactly what this means)
• The repeating correlation coefficient of sequencing technology is above 0.99. High repetitive correlation means that there will generally be no problem when carrying out downstream analysis.

This was done because they were not able to generate the number of reads promised in the contract. Although it sounds like there shouldn't be any glaring problems, I'm not sure how common this method is in the field.

Thanks.

[ATpoint]

The first point means that they regularily check the quality of their stuff (nothing special, every company should do that) and the second one means that the samples are (probably) highly comparable in terms of linear read count correlation, as expected for a technical replicate of the same RNA/cDNA. It is probably ok to simply merge the counts. If you want to investigate further, you can quantify the runs independently, log/normalize them, e.g. rlog or vst from DESeq2, and then use principal component analysis to see if the technical reps cluster together (which is quiet unlikely given that they did not mess up the cDNA prior to sequencing by DNase contamination or anything, which is again, very unlikely).