Question: More Problems with Cuffdiff
gravatar for gkuffel22
3.2 years ago by
United States
gkuffel22170 wrote:

Hope someone can help,

I have some RNASeq data and I have 2 groups (Control and Treatment) each containing 3 replicates. When I run Cuffdiff just as proof of concept I ran the groups (Control and Treatment) with one sample in each group. This worked great, the gene names were assigned and there were values present for log2(fold_change) everything looked good. Now, when I add my replicates to each respective group 3 Control against 3 Treatment Cuffdiff produces zeros across the board, no log2 fold change, no FPKM....I still get the gene names but it's as if Cuffdiff couldn't perform any calculations. Please don't say this is an issue with my GTF file matching my genome FASTA because these files work perfectly when only using 1 sample per group. Please help if you have any advice.

cuffdiff • 967 views
ADD COMMENTlink written 3.2 years ago by gkuffel22170
gravatar for Jennifer Hillman Jackson
3.2 years ago by
United States
Jennifer Hillman Jackson25k wrote:

Hello, Is this reproducible on Or was this work done there? If so, please share a history link and email that and a link for this question to Thanks! Jen, Galaxy team

ADD COMMENTlink written 3.2 years ago by Jennifer Hillman Jackson25k

Hi Jen,

Thank you for taking the time to help me. This work was initially done on our local instance of Galaxy but I reproduced the work on and the result is the same. Here is a history link:

ADD REPLYlink written 3.2 years ago by gkuffel22170

Thanks for sharing the history. I do not see anything wrong with the inputs. Instead I believe this is a content issue. I did notice that the differential expression testing is coming up as "NO TEST" in the result files for both the single and multiple sample runs, it is just sparser in the single run (some data does pass through).

This "NO TEST" result is explained on the Cuffdiff manual linked from the tool form. In short, it means that the data does not meet the minimum requirement by the chosen parameters to perform the analysis. I would start by reviewing the alignments to confirm that those were done in an optimal way plus to get an idea of coverage. Then once you know that information, experiment with the Cuffdiff parameters themselves so that they are a better fit for your data. Take care, Jen, Galaxy team

ADD REPLYlink written 3.2 years ago by Jennifer Hillman Jackson25k

Hi Jen,

I also have some RNAseq data with two groups (Exposed and Control). The exposed contain 3 replicates and the Control were done in duplicate. After running Cuffdiff on, a good majority of the reads generated read "NO TEST", and I only discovered about 24 statistically significant differentially-expressed genes. I would have imagined to have more significant DEGs. I believe my alignments are good. Can I provide you with my history and you can take a look? Would I be better off using DeSEQ2 instead of Cuffdiff? I am relatively new to analyzing RNAseq data so I do not know what would be best.

I appreciate your help!



ADD REPLYlink written 2.7 years ago by mrm6120

Hello Matt,

Maybe test out DeSEQ2 and compare? If there are significant differences, then this points to issues with the Cuffdiff parameters fitting the data: coverage too shallow/sparse, coverage overly deep, and the like. Have you visualized the Tophat2 results for a few regions with a NO TEST result across the samples/replicates - visualizing almost always helps with context when hunting for the root cause of an unexpected result. I am assuming that you already checked the Tophat2 mapping rates and these are acceptable - if very low, double check that the fastq data mapped is actually in fastq sanger format.

Above are the same items I would be checking. But if after all this, you'd still like me to take a quick look, please share a history link. Either here (warning, anyone would be able to view it), or send it to along with a link to this post in the comments.

Next time it would be best to post a new question as a new thread. But it is OK this time, these are related and let's get you the help you need without fussing about where the post is.

Thanks, Jen, Galaxy team

ADD REPLYlink written 2.7 years ago by Jennifer Hillman Jackson25k
Please log in to add an answer.


Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 16.09
Traffic: 169 users visited in the last hour