The correlations seen in the the genome-wide-association study are impressive, but I don't see how they would help in this case. If you'd use that data for a mo

The correlations seen in the the genome-wide-association study are impressive, but I don't see how they would help in this case. If you'd use that data for a model, you'd end up with a model that has thousands of parameters, and each of them has a huge margin of error. Using this data, how precise would your model be? Ie. given a genome, how precisely could you estimate height? Concerning your comparison with breeding: you seem to misunderstand me. I don't think that breeding per se doesn't work. Breeding by looking at phenotypes works wonderful. But what you suggest is to just iterate to optimise [according to my spellchecker that is how you spell it] according to a model. You are missing the step where those embryos grow (or fail to grow) into adults. Without this step, your optimisation procedure will drift randomly on every axis that your model doesn't measure.