Like, Levine’s regressions (not PRS, but w/e) often have drastically different coefficient values depending on dataset [and due to interaction effects, assigning high coefficient to one value often decreases coefficients on other values if you compare some regressions w/each other]
There are PRS scores on intelligence (the new thiel fellow genome sequencing startup tries this) and longevity [anton kulaga and olga have their own longevity PRS models and they give you a t-shirt showing your longevity PRS but I suspect this is highly underdeveloped]
