These pathways were ranked as follows in a reputable journal publishing an article about epilepsy related-genes, with -log (p-value) in bold; followed by p values in parentheses:
1. (top rank) Molecular Mechanisms of Cancer 9.66 ( 2.188 x 10-10)
8. Type II Diabetes Mellitus Signaling 6.14 (7.244 x 10-7)
11. Pancreatic Adenocarcinoma 5.7 (1.995 x 10-6)
A plausible pathway related to the Central Nervous System (CNS) appeared lower down:
12. Axonal Guidance Signaling 5.67 ( 2.13 x 10-6)
Does anyone else see highly ranked false positives in pathway analysis in their field- is this a systemic problem? Do you think it is a significant problem? How do you get around it? (It is possible to run several different pathway analyses, but they are often contradictory, and it should not be necessary to cherry-pick).
Do you think the problem lies in the realm of biological classification and annotation, and/or the algorithms used, and/or the statistics?
If this is an important issue, what do you think are potential solutions