Fig. 3

The impact of rare and novel noncoding variants on gene expression. a Using the replication set, we looked at the adjusted proportion of transcripts with common (red), rare (blue) or novel (green) noncoding variants in the vicinity (+/−20 kb) of a gene based on different allelic imbalance: 1.5 to 9, 2 to 9, 2.5 to 9, 3 to 9 and 3.5 to 9 fold difference. Adjustment was based on average number of SNPs used to calculate ASE at each ASE levels. b Enrichment of proportion of transcripts showing allelic imbalance (AI) with rare or novel variants in the vicinity of the gene compared to AI transcripts with common variants in vicinity of a gene. We looked at coding (histogram) vs noncoding variants as well as noncoding variants in DHS regions correlated with the promoters (Pearson correlation r > 0.5 to 0.9). In red are all transcripts where allelic imbalance was measured (allAI) and in blue are the transcripts for which the top associated SNP is homozygous in the sample (homAI). Linear regression slope for allAI = 0.015 (p-value = 0.0196) and homAI = 0.063 (p-value = 0.0024). Allelic imbalance genes are considered as > =2 fold between the alleles and equally expressed genes are < =1.5 fold. c Fold difference between proportions of AI transcripts with rare or novel variants in the vicinity compared to AI transcripts with common variants in the vicinity. Only including transcripts for which the top associated SNP is homozygous (homAI). We looked at coding (histogram) vs noncoding variants around the genes (+/−20 kb from gene) and in DHS regions correlated with the promoters (Pearson correlation r > 0.5 to 0.9). We compare different levels of allelically imbalanced transcripts from 1.5 fold to 3.5. all AI: AI transcripts comparing all transcripts for which ASE was measured and homAI: transcripts for which the top associated SNP that drives the association across samples is homozygous