Weird Fragment Length distribution multiQC

Question

[_deb]

I have several samples coming from an RNAseq experiment generated with Illumina sequencing (Pair-End sequencing of 151bp).

I used Salmon for indexing the de novo transcriptome and then I quantified in mapped-based mode against the created index. However, checking the output from MultiQC I noted a very weird thing. The fragment length distribution suddenly drops around 145-150bp and then goes up again...

Is there someone that could have an explanation for this behaviour?

Any putative explanation is welcome!

Answer 1

[Carlo Yague]

Thats weird... could it have something to do with the way Salmon deals with dovetail reads ? For instance, these reads could be excluded when the insert about the size of the read length (150), but not when the insert is shorter (<145). I wonder what would happen if you use option --allowDovetail with Salmon in mapping mode. My guess is that it would solve the issue.

[_deb]

Thanks a lot for your answer! In your opinion, could this affect the downstream analyses (e.g. DE analysis)?

[GenoMax]

Can you tell us if the option mentioned above mitigates your observation? Have you tried it? How many cycles were sequenced for your data (300 each way?)

[_deb]

Yes, I tried adding the option --allowDovetail, and it solved the problem. It seems a normal default behaviour of Salmon, so there is nothing to be too worried about apparently. Thank you again!

[GenoMax]

Please accept the answer by @Carlo to provide closure to this thread.