Table 1 Summary of the effects of variable power and winner’s curse on eQTL effect size estimates and their relationship with minor allele frequency
Median eQTL effect
Full DataAscertainment (validation)
Expression increasing21%22% (18%)
Expression decreasing−18%−18% (−14%)
MAF
0.02–0.1MAF
>0.1MAF
0.02–0.1MAF
>0.1
Expression increasing—filtered for power48%40%52% (44%)45% (45%)
Expression decreasing—filtered for power−28%−28%−29% (−25%)−29% (−29%)
  • Effect sizes reported as percent change of eQTL-heterozygotes relative to the homozygous ancestral. To explore the effects of winner’s curse, individuals were split into an eQTL ascertainment set and an effect-size validation set. To explore the impact of variable power to call eQTLs of the same effect across allele frequency bins, eQTLs were filtered to remove those with estimated effects that we would lack power to call at low allele frequencies. Here, we consider eQTLs with estimated effects larger than the minimum magnitude of estimated effect for significant, rare eQTLs (MAF<0.02). To ensure conservative estimates of the relationship between effect size and allele frequency, we also remove rare eQTLs (MAF<0.02) from comparisons across allele frequency bins.