How/when to use cuffmerge and/or cuffcompare?

Question

[awilli84]

Hello,

Completely new to RNA sequencing and bioinformatics and I have been trying to figure things out on my own so far...I am trying to analyze my RNA-seq data and I have made it as far as using cufflinks on all my files to quantify my transcripts. So now I have a bunch of transcripts.gtf files for each of my samples. I think I am supposed to use cuffmerge but I don't know which .gtfs to merge. Say I wanted to compare differential expression of my two controls (two different untreated cells types). Do I merge the .gtf files produced by cufflinks for these two samples then use that merged.gtf file for cuffdiff? Honestly, I'm lost and I'm having a hard time understanding why I wouldn't just compare the accepted_hits.bam files from tophat to a reference genome .gtf without using cufflinks. Also, why would cuffcompare be used? I am so new to this, the explanations on the cufflinks website are going over my head.

Any help greatly appreciated.

Ashley

Answer 1

[andrew.j.skelton73]

Check out the image on this page, which shows how the general workflow works. Cufflinks is a transcript assembler, so it'll go through your reads and try to figure out what's a potentially new transcript. So for samples A-Z, you run cufflinks on all of them individually, then merge the subsequent GTF files to get a complete transcriptome of all your samples. For example if sample 'A' only finds transcript 'X', then it'd be handy to quantify it in all your other samples too, so that's the point of cuffmerge. The whole point of using Cufflinks is to look for novel transcripts.