So you're saying we get basketball players and jockeys, then look at only the genes that "determine" height, then cluster them? I don't see how that would be useful at all. Predictive validity is the sine qua non of a scientific construct. What would this predict? Nothing beyond what you already knew. This is called Diamond's fallacy I think, after the Marxist charlatan Jared Diamond. He claimed you could make an "eye-color/shoe-size" race, or something. Which would predict nothing beyond the initial classification. However clustering based on overall genomic similarity would be highly predictive for any number of things. Do you make these same objections about "size classifications" being possible when people classify bears by shared ancestry too?