The problem with SNP-based association is that, for the vast majority of disease-associated SNPs, the relative risk is only increased on the order of a few percent at most. When it comes to predicting the risk of an individual (as opposed to a population) developing a particular disease this is virtually meaningless. No doubt the current technology will appeal to some out of curiosity. However I do think that the vast quantities of sequence that will be generated by next generation platforms over the next few years will lead to the ability to usefully predict disease susceptibility of an individual basis.