Heritability and genomics of gene expression in peripheral blood.

Transcriptome-wide estimates of heritability, based on n=2752 twins. (a) Manhattan plot of h2 P-values for the highest h2 transcript for each of 18,392 genes. The inset (showing PADI2) illustrates that the evidence for heritability is based on higher a correlation between MZ pairs (blue) than between DZ pairs (red). (b) Clustering of 777 genes with h2 q < 0.05. The most heritable genes belong to the cluster with lowest inter-gene correlation, but many significant genes belong to clusters with high inter-gene correlation. (c) Among 43,628 transcripts, the significant proportion (in terms of false discovery q-value) is dependent on mean transcript expression, increasing rapidly for transcripts above an approximate detection threshold (expression ≥ 3.584, determined as the 90th percentile of chrY expression in females).

Gene density and other predictors of heritability, using n=2616 paired co-twins and 18,392 genes. (a) Mean h2 (corrected for gene expression level) vs. density of protein coding genes per autosome, showing that heritability is considerably higher for gene-poor chromosomes. Plot symbol area is proportional to number of array genes per chromosome. (b) Histograms of the permuted enrichment z-statistics for two predictors listed in Table 2. Observed values (blue dots) are extreme compared to the permutations.

Apparent heritability and local IBD effects vs. true underlying distributions. (a) For the twin-based h2 estimates (n=2752, 8818 expressed genes shown), subtracting the effects of sampling variation produces an estimated true distribution (blue). Re-simulating from the fitted true assumed distribution closely approximates the observed h2 (black curve). (b) The analogous expressed-gene results for local IBD effect estimation. (c) Proportions of all 18,392 genes exceeding h2 thresholds for observed data and for the estimated “true” h2 distribution. The MuTHER study (n=856) reported many more extreme h2 values, but the observation is consistent with greater sampling variation due to smaller sample size. (d) The analogous figure using only expressed genes from both studies.

Comparison and replication of eQTL results. (a) Number of unique genes with evidence of local association (q < 0.01, SNP ± 1 Mb window of gene), depicted for published leukocyte eQTL studies (LCLs, monocytes, and PBLs), as well as subsampling of NTR data (PBLs) using only genotyped markers and moderate QC (n=2494, 43,628 transcripts examined). Sample sizes are corrected for the number of covariates used. The “NTR with final QC” value applies q<0.001. (b) Overlap of local eQTL findings with two other large blood studies, at q<0.01. (c) Number of unique genes with evidence (q<0.01) for distant (greater than 1Mb) association. The implausible non-monotone pattern for NTR on original expression values illustrates the importance of robust association methods. Using the final QC on NTR data and q<0.001 drops the number of distant eQTLs from over 800 to ~300. The results suggest that many distant associations remain to be discovered, but careful QC is essential. (d) Overlap of distant eQTL findings (q<0.001) with previous studies (within 1 Mb).

Properties of distant eQTLs. (a) 348 eQTLs (gene-SNP pairs) were significant (q < 0.001) and passed the QC procedures and, of these, 165 replicated (q < 0.1) in 1895 NESDA individuals. (b) The 304 SNPs in significant eQTLs were examined for overlap with regulatory features, including DNase/FAIRE and transfactor binding sites, using Variant Effect Predictor (version 2.8) of Ensembl. 54 Most features were not enriched, although the 3 SNPs annotated as 5′ UTR variants all overlap with regulatory features, representing a significant enrichment compared to the total 18.4% overlap of distant eQTL SNPs with regulatory features representing a significant enrichment compared to the total 18.4% overlap of distant eQTL SNPs with regulatory features. (c) The π1 value represents the estimated proportion of the transcriptome influenced by the 304 QC-passing SNPs in significant eQTLs. Across all significant bins the cumulative proportion is only ~3%. (d) A distant eQTL hotspot on chr19 was associated with the expression of 12 distant genes, and one local gene (MYO1F). The partial correlation graph suggests that MYO1F expression is independent of the expression of the other distant genes given the expression of the transcription factor SOX13.