Hi adrian, I used topGO to analyze my data and get the following results: GO.ID Annotated Significant Expected pvalue GO:0016746 123 3 7.13 6.70E-08 GO:0016747 105 3 6.08 6.00E-07 GO:0016757 281 13 16.28 1.50E-05 GO:0016408 15 0 0.87 5.00E-05 GO:0001071 547 32 31.69 7.80E-05 GO:0003700 547 32 31.69 7.80E-05 GO:0046527 89 6 5.16 0.00013 GO:0016759 25 0 1.45 0.00029 GO:0004518 112 1 6.49 0.00062 GO:0016758 209 12 12.11 0.00078    Do you notice 'GO:0016408' and 'GO:0016759'? The number of significant gene of these two GOIDs  is zero. In this case, why they are considered as significant? Can we simply remove them from the list?   A second question is that : there are many combinations of 'algorithm' and test 'statistics' and the results are not the same and maybe huge different. Which method should we chose for the analysis? datong wang [[alternative HTML version deleted]]