Estimating strength of polygenic selection with principal components analysis of spatial genetic variation

Principal components analysis on allele frequencies for 14 and 50 populations (from 1K Genomes and ALFRED databases) produced a factor accounting for over half of the variance, which indicates selection pressure on intelligence or genotypic IQ. Very high correlations between this factor and phenotypic IQ, educational achievement were observed (r>0.9 and r>0.8), also after partialling out GDP and the Human Development Index. Regression analysis was used to estimate a genotypic (predicted) IQ also for populations with missing data for phenotypic IQ. Socio-economic indicators (GDP and Human Development Index) failed to predict residuals, not providing evidence for the effects of environmental factors on intelligence. Another analysis revealed that the relationship between IQ and the genotypic factor was not mediated by race, implying that it exists at a finer resolution, a finding which in turn suggests selective pressures postdating sub-continental population splits. Genotypic height and IQ were inversely correlated but this correlation was mostly mediated by race. In at least two cases (Native Americans vs East Asians and Africans vs Papuans) genetic distance inferred from evolutionarily neutral genetic markers contrasts markedly with the resemblance observed for IQ and height increasing alleles. A principal component analysis on a random sample of 20 SNPs revealed two factors representing genetic relatedness due to migrations. However, the correlation between IQ and the intelligence PC was not mediated by them. In fact, the intelligence PC emerged as an even . CC-BY-NC-ND 4.0 International license peer-reviewed) is the author/funder. It is made available under a The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/008011 doi: bioRxiv preprint first posted online Aug. 14, 2014;