Here I am exploring a way to represent the expression of this pathway - CI or se bars or similar. I am first scaling the data so that each gene will be eqully weighted when averaging the pathway expression and so that when plotting the 0 line will have been the mean expression across all samples. Before I squish the genes together I just want to explore a few different plots to orientate myself… It is all a bit too complex like this to use as the base for a plot… although the last there gives a glimpse of what we might cut up to use (it's not quite correct). I have a script for calculating se and 95% CI intervals This could be the basis for a plot where we just cut the panels up using illustrator and paste them to a tree like structure or similar developmental graph like in the Miller paper. I would say that the tighter group of 4 genes is superior. Many more groups are significantly different form the mean usign these alone. I attach both plots glucoSDexprs(4 genes) and glucoSDexprsFull(6 genes) The presentation would be slightly clearer than the Miller paper. You can interpret the Y-scale as standard deviations of the data. So group H for instance has a mean of ~1 std deviation below the overall mean of all samples. Whilst the basr are the 95% confidence intervals. So where the bars do not cross 0 you can say this group is significantly above or below the mean of the whole group. I can produce slight variants on this Fig to make it easier for cutting i.e. without strips or enclosing boxes. And I can also replace A-B with proper label names. Source.