test for bio-differences with RNA-seq

[daysfoot]

The test of the differences between two biology replicate based on RNA-Seq data is really bad.

By cuffdiff, there are 5600 different significant expressed genes in gene_exp.diff.

I can not figure out the reason?

Do you have the same problem ever?

[dpryan]

It sounds like you're wondering if one of your samples is an outlier. It's usually easiest to tell that sort of thing in a PCA graph (though you'll often have an idea beforehand when you do QC).

[chadn737]

Why is that bad? Depending on what the samples are, I do not find that unusual at all.

[GenoMax]

Wording in original posting seems to indicate that this may be a comparison of two (single) biological samples (replicates). But I may be wrong.

[chadn737]

Wording in original posting seems to indicate that this may be a comparison of two (single) biological samples (replicates). But I may be wrong.

If that is the case, then I would not consider looking for DE, especially by cuffdiff, to be at all a valid means of determining how similar/dissimilar the samples are. Particularly if the samples are not genetically homogeneous, etc.