What is the difference between cufflinks and cuffquant.
They are both used to quantify the expression of the transcripts.
My understanding is cuffquant uses the merged.gtf from cuffmerge for the samples that we want to compare whereas cufflinks uses the online annotation file. However, cufflinks can also use the merged file to quantify expression.
Cuffquant is simply an intermediate step that does the quantification, but with the additional benefit of quantifying all "novel" transcripts identified within all samples after CuffMerge. For example sample A finds novel transcript x, sample B finds novel transcript y - x is not quantified in B, and y is not quantified in A. Cuffquant allows this quantification using the merged GTF from cuffmerge.
edit: Check out the image on this page to get your head around what's happening in the intended workflow
Hi Andrews. Thanks for your reply. I had already checked the pipeline in cufflinks webpage. Just both cufflinks and cuffquant compute the gene and transcript expression profiles. And cufflinks has also the benefit to output novel genes and transcripts (check -g argument here). So I don't quite understand what the difference in the quantification of the transcripts if I run:
ADD REPLY
• link
updated 4.4 years ago by
Ram
43k
•
written 8.5 years ago by
aleka
▴
110
0
Entering edit mode
In principle, the quantification method should be the same, however if you give cufflinks the merged GTF and tell it to look for novel stuff, then it'll find more stuff - Which will be even more irrelevant. In my experience 99% of what Cufflinks calls "novel" is complete garbage. I'd honestly rethink your approach to analysis, because the trade off of "novel discovery" just isn't worth it. To get back to your question, Cuffquant will strictly do quantification without transcriptome assembly, so that's a big overhead to get out of the way, also the output is streamed into binary cxb files, which are optimised to be fed into CuffDiff.
Great. Now I have a much better idea. However I was wondering how you judge that the novel transcripts that cufflinks detects are "garbage"?
ADD REPLY
• link
updated 4.4 years ago by
Ram
43k
•
written 8.5 years ago by
aleka
▴
110
0
Entering edit mode
Take the top results from Cuffdiff and check the alignments and their respective coverages in IGV, see if it convinces you. Some will look convincing, which you'll have to confirm through qPCR, but others will most likely not convince you, which makes visual inspection a key element.
Hi Andrews. Thanks for your reply. I had already checked the pipeline in cufflinks webpage. Just both cufflinks and cuffquant compute the gene and transcript expression profiles. And cufflinks has also the benefit to output novel genes and transcripts (check
-gargument here). So I don't quite understand what the difference in the quantification of the transcripts if I run:or
where
merged.gtfcame from cuffmerge.In principle, the quantification method should be the same, however if you give cufflinks the merged GTF and tell it to look for novel stuff, then it'll find more stuff - Which will be even more irrelevant. In my experience 99% of what Cufflinks calls "novel" is complete garbage. I'd honestly rethink your approach to analysis, because the trade off of "novel discovery" just isn't worth it. To get back to your question, Cuffquant will strictly do quantification without transcriptome assembly, so that's a big overhead to get out of the way, also the output is streamed into binary cxb files, which are optimised to be fed into CuffDiff.
Great. Now I have a much better idea. However I was wondering how you judge that the novel transcripts that cufflinks detects are "garbage"?
Take the top results from Cuffdiff and check the alignments and their respective coverages in IGV, see if it convinces you. Some will look convincing, which you'll have to confirm through qPCR, but others will most likely not convince you, which makes visual inspection a key element.