Fig. 4

ASM is driven by allele-specific CTCF and TF binding in both normal and neoplastic cells. a X-Y plots showing examples of TF motifs with strong correlations between predicted allele-specific binding site affinities (PWM scores) and methylation differences across all occurrences showing ASM. These examples are among 179 significantly correlated motifs (Additional file 9: Table S8). Each data point represents one occurrence of the motif overlapping an ASM index SNP in cancer (orange) or non-cancer samples (blue). For occurrences showing ASM in multiple samples, allelic methylation differences were averaged across samples by sample type. R² and B-H corrected p values (FDR) were calculated using linear regression. b Most of the polymorphic motifs with significant correlations between allelic methylation and predicted binding affinities are also enriched among ASM regions (Additional file 10: Table S9). The heatmap shows the enrichment or depletion, in log²(OR), for the top 10 enriched TF binding motifs among cancer or non-cancer ASM loci in regions defined as chromatin desert or non-desert (the “Materials and methods” section). c Significant correlations between allelic TF binding affinity scores and ASM in each of the 4 classes of ASM loci. The graph shows the fitted ASM difference on PWM score using a multivariate mixed model. The fitted line and its 95-confidence intervals (area) are shown for each ASM class; slopes were calculated by the marginal effects of the interaction term between PWM score and ASM class and were significantly different from zero. The correlations are similar in cancer ASM (in both non-desert and desert) compared to non-cancer ASM, with only small differences in the slopes for each class