3
$\begingroup$

I'm using cuffdiff 2.1.1 to look for differential gene expression between two conditions. Each condition has 2 biological replicates. The results I get look promising from a log fold change perspective - lots of genes that I would expect to be affected have the highest fold change values. However, from a q-value perspective very few of the results look promising, having q values > 0.05.

My question is, given 2 replicates, how is the q value calculated? Surely if there was high variance between replicates but the difference between experimental conditions was also large then the q-values would be correspondingly high?

In this case, can I, or should I, really trust the q-values?

$\endgroup$
  • 1
    $\begingroup$ cufflinks.cbcb.umd.edu/manual.html#gene_exp_diff $\endgroup$ – fugu Aug 23 '13 at 12:58
  • $\begingroup$ If after a few days you don't get good answers here, consider reasking the question here: biostars.org It's basically stackexchange specifically for bioinformatics. About a third of the questions over there are about next generation sequencing. $\endgroup$ – swbarnes2 Jan 9 '14 at 18:01
3
$\begingroup$

q-value is the corrected p-value to account for multiple testing (i.e. you are testing thousands of genes). Those with q-value <0.05 are significant experiment-wise. Cuffdiff uses Benjamini-Hochberg correction to compute FDR (i.e. q-value). The calculation does not depend on the number of replicates it's based on the distribution of p-values those, yes, are influenced by the number of replicates. The more replicates you have the better).

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.