basic concepts behind --sjdbOverhang (STAR)

Question

[serpalma.v]

Dear community, after reading the manual and several threads (google: --sjdbOverhang), I still do not feel confortable using this option.

I understand the instructions for --sjdbOverhang as that it sets the max number of bp on each side of an exon-exon junction. And that is determined by the max length of a read. So, if a read has 50bp, the max number of bp for the overhang is 49, because at least 1 bp is needed on the other side.

In the manual it is stated that "the default value of 100 will work as well as the ideal value". In this case, the ideal value is 49; so, if the option is set to 100, my read would only use 49 out 100 bp of the overhang.

Is this reasoning correct? if yes: - is it then the default correctly set for any read of lenght <= 101?

I have search for an aswer for these question for the las 1.5 hout without succes, if I miss a thread, please let me know.

Thanks

EDIT: Addiotionaly, why the option --sjdbOverhang is included during genome indexing? it seems to be an option only applicable only if reads are being mapped.

Answer 1

[John]

Why dont you try running STAR with --sjdbOverhang 49 and then repeat with --sjdbOverhang 100, and check the difference :)

I'm not trying to be cute about it - I assume for reads <100bp that 100 will work fine as per your reasoning. However, that assumption assumes the STAR manual reflects reality, which, you know, who knows. Empirical evidence ftw.

[serpalma.v]

Thanks John, I uderstand your point. I could try that, but I need to understand what I am doing, so that later I can fully trust my results.

[John]

Right well, if you really want to understand it, you can search the STAR source code for all mentions of sjdbOverhang with this: https://github.com/alexdobin/STAR/search?p=1&q=sjdboverhang&utf8=%E2%9C%93

I don't personally understand c++ (else I would have read it and answered the question definitely), but maybe you or someone at your place of work can?