NovaSeq uneven lib sizes: any tagwise bias?

[David [R]]

Hi,
I just got my first NovaSeq-produced RNASeq dataset and the obtained lib sizes are quite uneven. The attached file shows the lib sizes. If you look at it, you’ll see that there are three libs between 75 and 100 M reads, another one that goes beyond the 100 M mark and, “best” of it, a library that shows ~400M reads (!!). On the lower side, one library barely gets 20M reads. The second plot in the pdf shows the number of mapped reads, reflecting the overall lib size tendencies.

The Sequencing Facility is not able to produce any plausible explanation to that 400 vs 20 M lib size range, provided that the pool behaved correctly in the MiSeq checking run. Have you ever encountered this problem? Is there any known tagwise bias in the NovaSeq?

As far as the analysis of the data is concerned, I know edgeR and DESeq2 packages deal with lib size differences but is not this too much?

I’d be grateful if you could help me out on this.
Best,
David

[nucacidhunter]

This result could be due to following:

1- Uneven pooling of libraries for sequencing
2- Differences in fragment size distribution of libraries (libraries with average shorter fragments will be sequenced favorably)
3- Presence of adapter-dimers in outlier library as they will be sequenced very efficiently

[David [R]]

Thanks a lot for your reply Jafar,

As far as pooling goes, apparently it gave good results in the MiSeq. As for primer dimers, correct me if I a wrong, if this was the case those reads would have not been mapped and ,therefore, that library would have seen dramatically decreased it percentage of mapped reads,. Right?
Lastly, I do not have access to the size distribution of the libraries, so I cannot tell about it. In any case, yet again, should not the same bias be observed in the MiSeq run?
Thanks again
David

[Bukowski]

In any case, yet again, should not the same bias be observed in the MiSeq run?

Not necessarily, the clustering of the flowcells is completely different on a NovaSeq to a MiSeq - so if they think the pool is balanced on the MiSeq, I'd certainly be looking at the size distributions of the individual libraries as this has always had a more pronounced effect on patterend flowcells.

[David [R]]

Thanks a lot.