running cufflinks with -G option result in novel transcripts

Question

[fta.mirzadeh]

Hi everyone! I'm running tophat2 with a reference annotation, then cufflinks with -G option providing the same annotation. using -G option I expect the result to be only about the known transcripts. however when analyzing with cummRbund I can find new isoforms for a gene. it is weired because I think cufflinks was not suppose to assemble any new transcripts while using -G option. how can I get rid of those new isoforms? is it possible to somehow filter the result in cummRbund?

[Rox]

It might not totally answer your question, but as I recently had a lot of trouble with Cufflinks, I can suggest you an alternative to Cufflinks ( that is much faster by the way) : Stringtie ( https://ccb.jhu.edu/software/stringtie/ ) I was stucked with cufflinks during days, and Stringtie solved my issue in only 10 minutes (I was also using tophat2 before) If you have some times and you are curious, maybe you should try and compare the result stringtie produce compared to the cufflinks results !

Hope it could help !

[fta.mirzadeh]

thanks I will try it out.

Answer 1

[Satyajeet Khare]

You can quantitate against the reference gtf file by directly performing cuffdiff with bam files generated by tophat2 (instead of using merged.gtf).

[fta.mirzadeh]

I've tried this before. I used my annotation.gtf instead of merge.gtf along the bam files. however while analyzing with cummRbund I found out that the software can not tell which isoforms belong to which gene. in other words isoforms were reported as a single gene.

[Satyajeet Khare]

It may have happened with gtf file (from UCSC?) in which gene ids and transcript ids are same. Try GENCODE gtf file. Also, --transcriptome-only option in tophat2 with your existing gtf file followed by cufflinks with default setting might do the job.