Hi Alicia: Thanks for your detailed reply! I have another question for the goseq method as stated below. I notice in the paper (http://genomebiology.com/2010/11/2/R14) of Young et al. (2010) that, for the category GO:0016020 "Membrane", it ranks 1st using GOseq for total read counts adjustment (Table 4), while it ranks 702nd using GOseq for length bias adjustment (Table 1). Why there is such an evident discrepancy between the two scenarios for different "bias.data" arguments? In the vignette of the goseq package, page 22, the top-6 categories using read counts adjustment match 3 categories of the top-6 categories when adjusting length bias, which I think is reasonable. Is there any rule of thumb that we consider top XXX (say, 100 or more) enriched categories as being kind of "equally important" so we don't really care their absolute rankings? If I am correct, those top-ranked categories turn out to be pretty "general" and they are at the very top of the GO hierarchy, which may be of little biological interests (for example, all CC, BP and MF are included). Probably people are more interested in finding more "specific" categories? Thank you in advance for your clarification! Best, Gu Mi -- Ph.D. student Department of Statistics Oregon State University Corvallis, OR 97331 USA [[alternative HTML version deleted]]