The rankProduct package recently posted on BioC has the following error: If you use rankProduct.oneChannel() and ((sum(cl==0)>1) && (sum(cl==1)>1)), then the FDR estimates returned are probably too low. Since another more robust and fully featured package implementing the rank product method will be posted on BioC soon, rather than fixing rankProduct I will ask that it be removed. I'm very sorry for any difficulties or wasted time this error may cause, and I'm very very sorry if you submitted a paper based on any erroneous data from my package. Sheepishly, -Ben