HISAT XS attribute

[timless]

Hi,

I've been aligning some unstranded RNAseq data using HISAT, then using cuffdiff to create fpkm tables.

I've encountered a problem where cuffdiff complains that the XS attribute is not present for spliced transcripts, but looking at my bam files there are many lines that contain XS tags.

I still end up with FPKM tables, and they look ok. But can I trust them?

I ran the same data through tophat then cuffdiff and got no errors from cuffdiff.

Any suggestions welcomed.

Thanks,
Tim

[Brian Bushnell]

I have encountered nothing but trouble with the Tuxedo pipeline and suggest you try alternatives such as DESeq. The XS tag is not part of the sam specification; it's very confusing, and the fact that CuffLinks (etc) requires it is just evidence that the tool is half-baked.

[Daehwan Kim]

In HISAT, XS attributes are not reported for alignments involving non-GT/AG splice sites. You can direct HISAT not to output such alignments using "--pen-noncansplice 1000000".

Thanks,
Daehwan

[timless]

Thanks for your help Daehwan, is there any way to get HISAT to put XS attributes on non-canonical splice junctions?

Do you think having splice junctions without the XS tag will have affected the results generated by cuffdiff?

Thanks,
Tim

[Daehwan Kim]

Hi Tim,

This version of HISAT doesn't have a way to output correct XS attributes for non-canonical splice sites, but hopefully I'll do some work to solve that problem in the near future. I think your tests with or w/o that option shouldn't affect differential analysis for transcripts involving canonical splice sites, but please feel free to let me know if you encounter something unexpected in your analysis.

Thanks,
Daehwan

[timless]

Ok, thanks again for your help Daehwan.