Running cufflinks without the -G or the -g option

[yl01]

Hi,

I am analyzing an RNA-seq dataset using tophat and cufflinks. Mostly I followed the procedures described in the Nature Protocol paper (http://www.ncbi.nlm.nih.gov/pubmed/22383036). According to the paper, in the cufflinks step to assemble transcript for each sample, the cufflinks was run without either the -G or the -g option. Because I wannted to find novel transcript isoforms, I shouldn't use -G (no novel transcripts assembly). I am just wondering if I should use the option -g (does an RABT assembly) or not. Actually the question is if cufflinks does novel transcript assembly if I don't use either the -G or the -g option.

Thanks in advance!

[413fei]

I have the same question. Can anyone answer this question?

[WhatsOEver]

Actually the question is if cufflinks does novel transcript assembly if I don't use either the -G or the -g option.

Sorry if I misunderstand the question, but if you don't provide an annotation file, how should cufflinks decide whether a prediction is novel or not?
However, you can still compare your assembled transcripts against an existing annotation afterwards using cuffcompare with the the -r parameter.

[Sheree Yau]

I'm also just starting to use cufflinks following the Nature Protocols paper. In the paper they provide a genes.gtf file at the cuffmerge stage, which as WhatsOEver says, is implementing cuffcompare to compare the assembled transcripts to an existing annotation. This way you can see which are novel transcripts and which were the same as predicted.

Related to this query.
My understanding from the manual was that when running cufflinks with -g genes.gtf "Output will include all reference transcripts as well as any novel genes and isoforms that are assembled" so -g should not stop isoform detection. However, in the Nature Protocol, -g genes.gtf is not used in the cufflinks stage. I wonder how including the annotation might affect isoform detection, for example, will this reduce sensitivity? or will this reduce false positives?

I have run cufflinks with the annotation and from my cuffdiff output, I seem to have some isoforms detected, but they are not significantly different between samples. I suspect it depends on sequencing depth, the organism and other factors but I'd be interested to hear the advantages or disadvantages of including the annotation in the cufflinks assembly stage too.

[yl01]

Thanks for the answers. I read a little more and seem to have the answer to my question. Without both the -G and -g options when running cufflinks, cufflinks will run using the riginal cufflinks assembler, which is able to detect novel genes and isoforms. Using the -g option, the RABT assembler will be used, which is also able to detect novel genes and isoforms. The RABT method is described in paper (http://www.ncbi.nlm.nih.gov/pubmed/21697122) in more detail, including comparing to the original cufflinks assembler. Still not sure which one is better though. Practically the running by using RABT takes significantly longer.