Detecting different SJ when using R1 only versus R1 & R2

[njoye26]

Dear Alex,

I'm running STAR on TruSeq RNA Access 2x100 fastqs that were q30 filtered and trimmed by Trimmomatic for over represented sequences detected by FASTQC. I'm detecting different SJs when I run STAR using just forward, paired reads versus forward and reversed, paired reads (ie, --readFilesIn *_f_paired.fq verus --readFilesIn *_f_paired.fq *_r_paired.fq). Sometimes the first base of the intron between the two outputs are the same but the last base of the intron differs. Ideas why I'm losing introns when I use all of the paired end data instead of half? Shouldn't the position of the SJ also be detected within a single read?

Thanks,
N

[Alexander Dobin]

Hi N

for PE runs with default parameters, STAR requires that both mates map at least partially.

Because of this requirement, some of the reads that map as SE, will not be mapped at all as PE, and some may be mapped to different loci. This will result in some junctions discovered only in SE alignments.

If you need to recover such junctions, I would recommend first mapping the reads as PE, and then re-map unmapped reads as SE.

Cheers

Alex