To further assess the false-discovery rate (FDR) for local and distant eQTLs, we divided the MRCA sample into two independent data sets with approximately 200 subjects per data set. One data set was used for discovery; the other, for replication. For multiple significance thresholds, we calculated the proportion of discovered eQTLs that could be replicated in the second data set with LOD > 3 (P-value of < ∼2 × 10−4) (Fig. 2). Local and distant syntenic effects had higher replication rates even when using low thresholds for significance, confirming prior observations that local eQTLs are more likely to represent true associations. The PC-adjusted analysis gave higher replication rates for both local and distant eQTLs, indicating that PC-improved analysis helps define reliable distant eQTLs (note that the replication rate here is likely underestimated due to sample size limitation) (Ding et al. 2010).
