Does Cuffmerge discard single-exon transcripts?

Question

[colin.kern]

I'm using cuffmerge to combine the transcripts built using cufflinks on different samples. I've just noticed that the annotation I get from cuffmerge contains no transcripts with only a single exon, they all have at least 2. Looking at the cufflinks output for the individual libraries, they do contain transcripts with a single exon. Does cuffmerge just automatically discard any single-exon transcripts, or there some other metric it uses and it just happens that none of the single-exon transcripts pass the metric? I don't see any options for cuffmerge to adjust this behavior.

Answer 1

[Satyajeet Khare]

cuffmerge does not exclude single exon transcripts. You can check canonical histone genes which are mostly mono-exonic. In merged.gtf file, they should have only one exon marked as "exon 1".

[colin.kern]

You are correct, those genes are there. I forgot to mention that I have filtered out all transcripts with class_code "=" since I'm interested in finding novel long non-coding RNA. This leaves me with no single-exon transcripts. Could this be because cuffmerge "automatically filters a number of transfrags that are probably artfifacts"?

[Satyajeet Khare]

Yes, cuffmerge may have filtered them out. Btw, many long non-coding RNAs are polyexonic. And therefore you may find novel lncRNAs in your sample.