Perhaps the single best piece of evidence in support of racially genetic differences in IQ is the Minnesota Transracial Adoption study which was interpreted by Richard Lynn (in his 2006 book Race Differences in Intelligence: An Evolutionary Analysis) as showing that the U.S. black-white 15 point IQ gap is 100% genetic, because in this study, even when whites and blacks are both adopted into white upper class homes, more than the full IQ gap remained.
However as commenter Race Realist reminds me, recently the study has been undermined by British physicist Drew Thomas who argued that one reason adopted whites scored so much higher than adopted blacks is that the white sample suffered from attrition. Thomas writes:
A total of 25 White adoptees were in the study when it began, nine of whom were lost at follow-up. The lost adoptees had relatively low IQs, so the remaining White adoptees were unrepresentatively high in IQ, as Mackintosh observed . One can prove this by comparing the original IQs of the full sample and the subgroup who were measured at both ages 7 and 17; the latter subgroup had an initial mean IQ of 117.6 (with a minimum IQ of 92) but the full sample had an initial mean of 111.5 (minimum 62). Because initial and final IQs had a correlation of 0.63 among the White group, the elite subgroup would likely have had their final mean IQ inflated by about 0.63 × (117.6 − 111.5) = 3.8 points. Meanwhile, the BW and Black–Black adoptees lost to follow-up hardly differed in IQ from the remaining adoptees, so attrition inflated those groups’ mean IQs by about only 0.2 and −0.7 points respectively. Adjusting the final mean IQs accordingly (Table 2) implies smaller racial differences of 3.5 points (White vs. BW adoptees) and 11.7 points (White vs. Black–Black adoptees) in the study’s final wave.