Salmon GC bias with caution

Question

[Morris_Chair]

Hello everyone, I'm using salmon for the alignment of SE sequencing fastq file. at some point I have this output:

[warning] Fragment GC bias correction is currently *experimental* in single-end libraries.  Please use this option with caution.

It's not clear to me what does it mean to use this parameter ( --gcBias ) with caution, what happens if I use it? can I or not?

Thank you, in particular to who made this amazing tool

Answer 1

[h.mon]

There is a lengthy discussion about it at Salmon issue 83:

"--gcBias" option for single-end reads?

The crux of the problem is (and I quote rob-p, Salmon developer) "Given a single-end read, one does not know the length of the fragment from which it originates". If I am not mistaken, the method implemented just uses the mean fragment length to apply the GC-bias correction, and wasn't subjected to extensive validation, thus it is considered experimental.