In both the current study and the Sinnott and Kraft (Sinnott and Kraft 2012) study, SNPs with R2 >0.3 but exhibiting bias were predominantly SNPs with lower MAF (≤10 %). It is also the case that both studies used HapMap reference panels for imputation. Since 1000 Genomes panels are enriched for lower MAF SNPs, one would expect that imputation based on 1000 Genomes panels would generate greater rates of bias than observed in these studies. Thus, as the field moves forward to take advantage of these more comprehensive reference panels for imputation, correction of this cross array imputation bias will be even more important.
