Chimeric.out.junction and SJ.out.tab

[Mario Keller]

Good morning,

I have an unstranded library (paired-end) that I mapped with STAR. Now I'm interested in the quantification of junctions and had a look at SJ.out.tab and Chimeric.out.junction. As the library is unstranded one does not actually know to which strand a junction should be assigned. However, it looks like STAR looks for motifs (e.g. GT/AG) to assign a strand (if possible), which is pretty cool.

For SJ.out.tab I have the number of uniquely/multi mapping reads supporting each junction. However, for the raw Chimeric.out.junction file this information is not present. What I do is that I aggregate rows, supporting the same chimeric junction (based on the first 9 columns). After aggregating identical rows I checked the output and I observed two rows that look pretty similar except that the acceptor and donor columns are flipped (1=4, 2=5, ...). Could it be that these rows reflect the same splice junction and should be aggregated as well?

Thanks in advance. 

[Alexander Dobin]

Hi Mario,

yes, these rows can be combined. Chimeric.out.junction reports the alignment of Read1 and reverse complemented Read2 without considering strand, so donor and acceptor designations will be flipping back and forth for unstranded libraries.