cufflinks "x" class code

[yfukuda]

Hi adarob, wenhuang,

Taking closer look at my sam file, I also found XS tags for reads on splicing-junction.
Should we remove or rewrite them..?
Thank you.

[wenhuang]

Hi Adam,

Glad to see this post brought your attention .

I used Tophat 1.1.1 and Cufflinks 0.9.2.

Mine is paired end reads. I don't see XS tags for unspliced reads but I see XS tags with strand information for junction reads.

My problem is that I see many single exon transcripts assembled by Cufflinks classified as "x".

I look forward to your reply.

Thank you!

[yfukuda]

Hi adarob,
Thank you for your comment.

I used tophat-1.0.3 (gangcai edited version to enable --best --strata options, http://seqanswers.com/forums/showthread.php?t=4099)

I don't find XS tags for reads.
However, results from edited version of tophat look slightly different from that from latest tophat (it produces two accepted_hit.sam files in the current directly and output directly. One contains mapped result and the other is for only tags). If tophat can affect this problem, can this have something to do?

Thank you.
y

[adarob]

Which version of TopHat did you use? I believe there was a bug at some point that may have caused this. If you look at your BAM file, are there XS tags for every read?

[yfukuda]

Hi,
Maybe I also have a same question.

I performed cufflinks and cuffcompare (with the latest version of 0.9.1) to analyze single-end, unstranded mRNA-seq, and I obtained a lot of 'x'-categorized isoforms, almost as many as '='.

$ grep "x" WTtranscripts.tmap |wc -l
4879
$ grep "=" WTtranscripts.tmap |wc -l
6042

I am confused about this 'x' because I think strand information of each isoform is unavailable from single-end, unstranded RNA-seq..? In this case should I ignore 'x' or should I sum up '=' and 'x' etc to count the total number?

Sorry if this question is stupid, but I appreciate someone's advice.
Thanks in advance.
yfukuda

[wenhuang]

Hi,

In the newest version of cufflinks, there is a "x" class code documented as "exonic overlap with reference on the opposite strand".

I have encountered many assembled single-exon transcripts that falled into this category. I have used unstranded RNA-Seq and have not provided any option specifying library type. I was wondering how cufflinks or cuffcompare, for this matter, determined the strand of a single exon transcript.

Thanks a lot for your help!

Wen