what does "replicates" mean in CuffDiff

[yub]

I am new to RNA-seq and want to find out how to use CuffDiff correctly:

I am trying to compare two groups or conditions (A vs B), with 5 samples in each group. Each sample is from a different individual.

After using Tophat-->Cufflinks for each sample-->CuffMerge, I tried to use CuffDiff to find differentially expressed genes between the two groups. When I input the 5 samples in each group as 5 replicates, CuffDiff generated an output where almost all tests were "FAIL". When I input each sample as a group without replicate, CuffDiff kept running without generating any result.

Thanks for letting me know what "replicates" mean in CUffDiff and if there is any solution to my problems!

[mbblack]

If you entered them as two comma separated lists (with a space between the comma separated input files for the two groups):

e.g. Group1_1.bam,Group1_2.bam,Group1_3.bam,Group1_4.bam,Group1_5.bam Group2_1.bam,Group2_2.bam,Group2_3.bam,Group2_4.bam,Group2_5.bam

Then they would be treated properly as biological replicates.

My understanding of the "FAIL" error is that it means the isoform abundance computation failed due to an actual numeric error (missing values can cause it for one, but there are other conditions that will cause it - search the board for related posts as there are discussions on that error in the past).

[yub]

Thank you very much, Dr. Black.

So one replicate just means one individual sample, therefore different replicates can be from different individuals, not necessarily from replications of the same individual. Is this correct?

How can I find out the variance or standard deviation in each group? CuffDiff output I got only had the mean of each group and p-value. Thank you!

[mbblack]

If you are talking about an experiment that used multiple individual subjects or animals, then yes, each individual would be a sample and each sample in each treatment group would be a biological replicate. Biological replicates sample the variation within a population.

A replicate taken from the same individual would be a form of technical replicate, which would tell you nothing about the biological variation within the population.

I'm not sure where you can actually get SD values. Check the cuffdiff manual and maybe see what is in the various FPKM tracking files - they may have what you need to compute it yourself. I have not used cuffdiff in awhile, so I don't remember the exact file contents of all the output files.