Fig. 3

Type II probe features impacting signal intensity and reliability. a Dependence of mean signal intensity (MI) on C content of the probe (for type I probes and other datasets see Supplementary Fig. S7). b Odds ratio for C enrichment in different DNA regions: Islands, due to their inherently high CG content, have a strong enrichment of probes with a large C content, which leads to the fact that on average MIs are higher. Odds ratios and associated p-values were obtained with a two-sided Fisher test (full results given in Additional file 2). Non-significant (FDR-adjusted p-value > 0.05) results are shown in grey. Significant results (FDR-adjusted p-value ≤ 0.05) with an odds ratio > 1 or ≤ 1 are shown in red and blue, respectively. c, d Differences in MI and unreliability scores depending on the CpG context of the probe target. e Dependence of MI (Y-axis) on the degree of non-unique overlap for 3’ nested subsequences (estimated copy number; X-axis) and C content of the probe sequence (color scale). Copy numbers for 10–50 bp long 3’ nested subsequences were calculated by Zhou et al. [14] by scanning 3΄-subsequence of the probe's source sequences (‘G's converted to ‘A's) and of the specified length in the bisulfite converted genome (GRCh38 assembly) for matches. f Reliability scores proposed by Sugden et al. [15] versus unreliability scores defined here. g Association of Reliability scores proposed by Sugden et al. [15] with SNPs and MI. h Methylation status (β) and intraclass correlation coefficient (ICC) for example SNP-associated CpGs and probes yielding low signal intensity of probes