Fig. 3
From: Revisiting genetic artifacts on DNA methylation microarrays exposes novel biological implications
Systematic evaluation of CpG/SBE–SNPs and probe-SNPs and examples of CpGs subject to genetic artifacts with additional levels of complexity. A Ambivalence in noise-to-signal ratio, BC(CVlogT), distribution across CpG/SBE-SNP categories (SNPs with MAF < 0.1 were excluded from this analysis). B Relative distributions of U/M plane cluster counts estimated by the K-caller across CpG/SBE-SNP categories (SNPs with a MAF < 0.3 were excluded from this analysis). C BC(CVlogT) of bi-allelic probe-SNPs as a function of distance to the 3′-end of the probe and targeted strand (SNPs with a MAF < 0.1 were excluded from this analysis). The obtained curves have been smoothed via local regression (loess). D Genetic artifacts on a CpG site associated with two contiguous SNPs and the estimation of haplotype frequencies from cluster counts in the U/M plot via general-purpose optimization. Reported haplotype frequencies were obtained from LDhap. E Genetic artifact caused by an indel that additionally interacts with imprinting. Amat and Apat indicate maternal and paternal alleles, respectively. Reported MAFs correspond to 1000 genomes (phase 3) of EUR ancestry (n = 1006). Allelic frequencies were estimated from cluster counts estimated via bivariate Gaussian mixture models from U/M plots and only concordant monozygotic twin pairs were taken into account in the computation. Twin-to-twin agreement matrices are also available on the plots employing the same color coding as the clusters in the U/M plot. Probe designs are highlighted in yellow boxes and SNPs/indels are denoted in bold red