RNA-Seq Star-Cuffdiff Pipeline error message after changing c parameter in cuffdiff

Question

[Aynur]

Dear Altruists, I am new to bioinformatics and please help me with this. I used Star-Cuffdiff pipeline without cufflink to get DEG analysis. When I run cuffdiff with default settings, I got gene expression data. When we try to validate highly expressed genes by qPCR, those genes are undetectable. So, my question is, where could be the problem? What can I do in this case? My library is truqseq fr firsstarnd , So I run cuffdiff again with library type , c flag, and b flag and then I keep receiving an error message saying labels should match the number of conditions. So, I retired again with the same code removing these advanced options, then it works fine. I have spent a lot of time on searching for this error, but could not get any useful information.
Please help me out. Thanks for taking the time to read.

Sincerely , Aynur

Answer 1

[i.sudbery]

Hi Aynur,

There are several things going on here.

First, RNAseq is more sensitive than qPCR, and so you can detect things by RNAseq that you can't by qPCR.

Second CuffDiff is old and deprecated software, which even the authors won't recommend you used these days for doing differential expression analysis.

There are several standard pipeline that are probably equally good for your purposes. The two most common are:

• STAR-featureCounts-edgeR/limma/DESeq2
• Salmon/kallisto-tximport-DESeq2/limma/edgeR.

Importantly for your case, both edgeR and DESeq2 have steps to filter out or reduce the salience of genes that have high fold changes but very low expression levels.

[Aynur]

Thank you very much. I really appreciate it. My question is I have two biological replicates, and I was told for these pipelines Its required to have at least 3 replicates. I read DESeq2 manual, and could not find about minimum 3 replicates. Am I misinformed? Please help me.

[i.sudbery]

These pipelines require replicates - that is you need to have more than one sample in each group. But they will run fine with only 2 replicates. You should be aware though that if you only have 2 replicates the results won't be so reliable, and some failure to validate might be expected.

[Aynur]

Gotcha. At the moment, I have to deal with 2 replicates each and 10 samples in total. Also, after STAR, I used HTSeq to get raw counts then planning for DESeq.