CompareGroupsGridFix pvalue

[TemenuzhkaDi]

In CompareGroupsGridFix would it be possible to have a heatmap with the pvalue (in log) ?
Also to control for the multiple hypothesis problem, would it be possible to have a correction of the pvalue, bonferroni or anything else?

[jozsarato]

hi, sure, these are possible.. though I think the t-value map heatmap is easier to interpret, than a p - value heatmap.. log p value is a good idea (i have never seen that kind of heatmap before).

[TemenuzhkaDi]

Great! For the heatmap with pvalue I think only squares with pvalue <0.05 need to be colored (the squares with >0.05 are not informative and can stay uncolored).
Do you think it would be possible to have 2 color gradients for pvalue depending when set A or set B is enriched. For instance, when set A is enriched pvalue 0.05-->0 goes from light yellow to red and when set B is enriched pvalue 0.05-->0 goes from light blue to dark blue.

[jozsarato]

hi, I do not think coloring by exact p value would be a good practice.. also this sharp cut-off at 0.05 is not really a good practice anymore, as it gives false confidence, and makes a big difference between P=0.06, and P=0.045, even though these imply quite similar results, so the visualization for them should also be similar

[jozsarato]

user should be able to select whether to plot t or p value based difference maps (maybe with bonferroni correction)

[jozsarato]

user can now select logP or t value based grid comparison visualization