Machine learning techniques for biomedical prediction

Apologies for cross posting,

I have a general question about why ML is not used more in translational research. I know that some/most bioinformaticians are aware of these approaches. Having now done some serious reading and got a bit of practice with WEKA, the relevant packages in R, adjusting bias and variance, had a good look at the Stanford lecture series, reading Hastie etc., I wonder why I haven't come across these techniques in the biomedical literature before. We often have vast numbers of parameters, clinical, demographic, psychometric and now biological (imaging or genetic etc.). ML techniques / rPART would seem perfect for generating knowledge from this info without spending 10 years in the molecular biology lab. CART in particular seems to have been used in big datasets on Cardiac disease and Leukemia and then ... nothing. Am I missing something significant regarding the weaknesses of these techniques or is there just a gulf between traditional statisticians who consult on clinical trials etc, and bioinformaticians who are occupied with GIWAS, omics sciences etc.?