Strolling with a skeptical biochemist
[Hat Tip: A Blog Around the Clock]
Microarrays are very good at picking noise and selling it as signal. It's not much better in other 'omicses. My personal favorite is "interactomics" - which we know almost all noise simply because there is virtually no overlap between results from different organisms or from different methods using the same organism. The very little there that is not noise all represents strong enough signals that were already known without any high throughput methods. Even in genomics it seems that different ways of identifying genes under positive selection produce very little overlap.