General Info About Cufflinks: What Is Teh Difference Between Cuffmerge And Cuffcompare? --Help

Question

[Ftorri]

Hi, sorry to bug with probably a naive question. I'm new to cufflinks. I am exploring all the tools. I have two main issues:

(1) I tried all the possible reference annotation files (refseq, UCSC, ensemble) but the output file of cufflinks (.gtf) doesn't have the gene_name even if for example the ensemble GTF files has it. SO my question is: how can I keep this info? Is it like that because you upload the file in UCSC without looking at it like a text file?

(2) If I have different samples (lets say brains of different people) and I want to evaluate the DE among them do I have to us ethe cuffmerge or the cuffcompare script?

Two examples:

-for testing I have one control brain and one affected brain: after having run cufflinks I have two transcipt.gtf. Do I have to use cuffcompare (to get a unique .gtf) and then cuffdiff?

-I will have four control brain and four affected brain, do I will have 8 transcripts.gtf files. Do I have to use cuffmerge grouping the 4 controls (producing one control.gtf file), the same for the cases (producing a cases.gtf). And then I have to use cuffcompare between the control.gtf and cases.gtf file having a unique .gtf file and then cuffdiff?

I don't understand if cuffmerge is to merge together ssample of the SAME groups and cuffcoimpare to unify different categories of samples..

thanks!!!

federica ftorri is online now Report Post Edit/Delete Message

[Bharat Iyengar]

perhaps this is quite late but still it may be useful for others who refer this post.

This page compares different strategies of producing merged assemblies with Cufflinks.

[Ftorri]

Yes, but if I have cases and controls how I can specify the group? Lets say that I have 5 cases and 5 controls, what is the path I have to follow?

-Run tophat for each sample having back 10 accepted_hits.bam files

..and then? I am missing something bc if I have ten samples , how can the software know that I want to compare the 5CA vs the 5 CTRLs?

thanks!

Answer 1

[Docroberson]

Cuffmerge merges the transcripts from all of the samples you input into it into one unified list of transcripts. You'll want this for differential expression. Cuffcompare will allow the comparison of that merged list to known genes (your gtf file).

(1) Run cuffmerge. Then run cuffdiff on the merged.gtf file using your annotation gtf. The output will include the accession ID (NOT SYMBOL) of known genes. You may want to run JUST cuffmerge with the --ref-gtf option to create a merged file with annotated genes, cutting out the middle man.

(2) Create your merged, unique gtf with cuffmerge and then run cuffdiff using the merged file as the input gtf. By default cuffdiff compares each group to each other group for every isoform in the merged gtf file.

Answer 2

[Paparountas Triantafyllos]

let the control be the first in the cuffdiff in the stated comparisons