Thank you Agnes, It makes sense to do it by the number of probes mapped to the GO instead of the number of hits on it. Thank you for the clear explanation. I only have one other question: Is the reason that molecular_function doesnt appear in the list in your first example that the probe had already been mapped to an end node below it in the list (i.e. binding, structural molecule activity, transporter activity)? I guess a more direct question is whether there are more annotations at MF in my second example because I left out a lot of the nodes under MF in that run (so the probes ended up contributing to MF and not the end nodes that were children of MF). Also, this is not directly related to my previous question, but do you think it makes sense to divide the number of annotations at each node by the total possible number of annotations at that node when using all 45101 probes on the chip? It seems like this would make it possible to compare the representation of different end nodes in your probe set to each other (since some GO terms have more children then others), but I feel like I may be missing something about comparing end nodes to each other. Thank you for the help, Dave