RNA-Star : SJ.out.tab differents according to 'outFilterType' value?? and

[ebe]

Hello,

we are using RNA-Star to our project (thank to authors!).
I don't understand a difference observed about the SJ.out.tab files (file with junctions) according to the value of 'outFilterType' option (Normal or BySJout).

Indeed, Alexander Dobin wrote (https://groups.google.com/forum/#!ms...w/VmgoGKSE9qQJ) :
"However, if you use --outFilterType BySJout option, splicing of alignments in the Aligned.out.sam will only be allowed across the junctions which pass the filtering into SJ.out.tab.
This option makes Aligment.out.sam file consistent with SJ.out.tab file."

So I had understood that the SJ.out.tab files should be the same, but not the sam files (file with alignments, from which the SJ.out.tab is created).
But I have two different SJ.out.tab files (all options except outFilterType are identical).
Could anyone explain to me why?
The other options are:
--outSJfilterReads unique --outSJfilterOverhangMin 29 12 12 12 --outFilterMultimapNmax 1 --outFilterMismatchNoverLmax 0.05 --outFilterIntronMotifs None --alignSJoverhangMin 5

Thanks!

[ebe]

The response from Alex Dobin, STAR's author :
using the --outFilterType Normal vs BySJout options will create slightly different SJ.out.tab counts for the following reason.
Imagine that you have a read two junctions, with only 1st junction passing the filter.
Then with the Normal option, the 1st junction will be counted in the SJ.out.tab.
If the BySJout option is used, the entire read alignment may be prohibited because of the 2nd junction, and then the 1st junction will be not counted in the SJ.out.tab.

An even more complicates situation may arise if after prohibiting the 3nd junction, the read maps into a different locus with a junction different from the 1st.

closed thread...