cuffdiff for between-group comparison

[shurjo]

Hello everyone,

I have been using cufflinks to get fpkm values and then analyzing these using the JMP statistical package for identifying between-group differential expression (8 cases and 8 controls). Can someone tell me if there is a way to do this using cuffdiff as well? From my reading of the manual, it seems that giving cuffdiff more than 2 sam files causes it to treat the data as a time series, which is not the case for my experiment. I considered just merging all the sam files from the 8 subjects in each group, but that way I will lose the power of the multiple replicates.

Any suggestions will be much appreciated,

Thanks,

Shurjo

[blackgore]

I had a similar problem recently, and had to try another program (DESeq) for dealing with groups - unfortunately cuffdiff didn't support the use of replicates. With luck that will change in a future release!

[xguo]

I had a similar problem recently, and had to try another program (DESeq) for dealing with groups - unfortunately cuffdiff didn't support the use of replicates. With luck that will change in a future release!

How do you guys test relative isoform abundance changes among multiple experiments? Is it easy to do hypothesis test based on Jensen-Shannon divergence in R or SAS?

[shurjo]

I had a similar problem recently, and had to try another program (DESeq) for dealing with groups - unfortunately cuffdiff didn't support the use of replicates. With luck that will change in a future release!

Thanks for your feedback! I am also using DESeq as an alternative.

[shurjo]

How do you guys test relative isoform abundance changes among multiple experiments? Is it easy to do hypothesis test based on Jensen-Shannon divergence in R or SAS?

I believe that Cufflinks has this functionality built in, using the algorithm that was published recently from the Wong lab (http://bioinformatics.oxfordjournals...reprint/btp113). However, I haven't gotten around to looking at this aspect of my data yet.

Thanks,

Shurjo

[xguo]

I believe that Cufflinks has this functionality built in, using the algorithm that was published recently from the Wong lab (http://bioinformatics.oxfordjournals...reprint/btp113). However, I haven't gotten around to looking at this aspect of my data yet.

Thanks,

Shurjo

cuffdiff extended the idea of using entropy to quantify relative isoform abundance and devised a hypothesis test based on Jensen-Shannon metric. However, as you said in the previous post, cuffdiff doesn't handle between-group comparison. Actually, it may be better to parse cuffdiff results and use other statistical package to do downstream analysis, which allows you more choices for differentially expression, clustering, and other things. I'm just not sure if there is any R package that handles Jensen-Shannon metric.

[blackgore]

it also wasn't the focus of my project at the time, so I didn't worry about it too much, I'm afraid.

[honey]

With my limited knowledge of stat I am trying to understand that DEseq basically do more quality control- normalisation etc and can also take replicates into consideration. Other than that if there are no replicates Cuffdiff/ Cuff comapre will provide the similar level of analysis. Any suggestion please.

Thanks

[Thomas Doktor]

Cuffdiff handles replicates now.