large variance of raw read numbers between samples

[mrfox]

Hi all,

My collaborators used Illumina HiSeq to sequence 5 samples (RNA-Seq). I am not sure how many lanes they used but the 5 samples were barcoded and sequenced with multiplexing. The 5 samples include 2 replicates for one condition and 3 replicates for the other. The read length is 50, single end. The samples were prepared and sequenced under the same condition. So I expect that we will obtain similar number of raw reads for the 5 samples.

However, I got 5.5G, 5.7G, 8.6G, 9.1G and 12G respectively for the 5 FASTQ files, reflecting the large variance of raw read numbers. Since the sequencing core would not provide quality control report or the detailed sequencing protocol, so I am not sure if this variance resulted from potential bias in some step of the sequencing experiment.

I wonder if anybody could tell if the observed read number variance is normal or it does reflect some bias.

THank you very much.

[mrfox]

Any advice? Thanks.

[dpryan]

You could look through the fastq files and likely determine how many lanes were used. They may also still contain the barcode information, so you could see how that was done. I've gotten a lot of variance in read numbers before in data returned from our various sequencing providers, usually having to do with the quality of the library before loading.

[kopi-o]

It looks completely normal to me (we do a lot of multiplexed RNA-seq at our facility)

[jgibbons1]

I'd say it looks pretty normal to me too. Obviously it's nice to have all samples in a tighter range, but it is pretty difficult to do so even with multiple modes of quantification. I recently multiplexed 8 samples, and quantified them with bioanalyzer, qubit and qPCR and still had a range of 25 million - 50 million reads per sample.