I have 10 samples which were treated with two different treatments. For these 10 samples, I have a measure to define that these 10 samples responded differently to the two treatments. Treatment A was active in 9 and inactive in 1 sample. Treatment B was active in 2 and inactive in 8 samples. Also I know the P53 gene mutation status in these 10 samples.
The problem is that I need to devise a statistical test that can tell me there's no/yes statistical difference in differential activity between treatmentA and treatmentB based on P53 mutation status.
Most simply, I would have done Fisher test to test the association of P53 mutation status with activity of treatment A or activity of treatment B and get 2 p-values accordingly.
I am confused if it makes sense to compute 1 p-value to tell me if P53 mutation status is associated with differential activity in the two treatments based on P53 mutation status.
Does anyone have any thoughts/suggestions?
Am I correct in understanding that these are the same cells being treated with both treatments? That is, a given cell can (or dish or whatever you're actually using) could show simultaneous response to both?
The 10 human samples are same, but underwent an expansion phase in mice to create xenografts. For 1st sample e.g., the mutation status of the two mice is exactly the same but one was treated with treatment A and another one with treatment B.
At least in theory one can do a 2x2x2 Fisher test. The question is if anyone has come up with a convenient implementation (there's a Fortran routine that can do it that was developed in the 90's, but that's probably not terribly useful).
Actually, I have solved my own problem. Thanks for your replies.
It'd be good if you mentioned your solution, should others run into the same questions.