using STAR .SJout files instead of bed2junc on .bam files

[sahin...@gmail.com]

Hi,

I've been using leafcutter downstream of STAR. In addition to outputting .bam, STAR also outputs a .SJout file, which contains a list of "high quality" splice junctions with read counts. It seems possible to use these .SJout files from STAR as the input to leafcutter clustering (i.e. instead of running bed2junc on the bam files and clustering based on that). Would you recommend one way or another? I have noticed that the resulting .junc file from bed2junc has many more junctions, sometimes up to 2x more, than the SJout file, presumably because STAR does some additional filtering step on junctions.

Thanks,

Sahin

[lhbsyl...@gmail.com]

hi,

may i ask you,which way you have choose ?  thank you!

在 2017年11月6日星期一 UTC+8上午5:35:00，sahin...@gmail.com写道：

[waqaskh...@gmail.com]

Hi, 

Did you get a answer for this query as I have also used STAR to generate sorted bam files using below parameters and I was thinking of using junction files to generate phenotype PSI matrix for sQTL analysis using FastQTL?

Parameters were: 

--sjdbOverhang 100
--outSAMprimaryFlag AllBestScore : output all alignments with the best score as primary alignments
--outFilterMismatchNmax 2/0(first/second pass): alignment will be output only if it has fewer mismatches than this value
--outSJfilterCountTotalMin 10 5 5 5 (non-canonical SJ and 3 canonical SJs)
--outSAMstrandField intronMotif
--outFilterIntronMotifs RemoveNoncanonical : filter out alignments with non-canonical junctions
--alignIntronMin 20 : mininum intron size
--alignIntronMax 6000 : maximum intron size
--outSAMtype BAM SortedByCoordinate: output sorted BAM file