High throughput characterization of genetic effects on DNA:protein binding and gene transcription

Abstract

Many variants associated with complex traits are in non-coding regions, and contribute to phenotypes by disrupting regulatory sequences. To characterize these variants, we developed a streamlined protocol for a high-throughput reporter assay, BiT-STARR-seq (Biallelic Targeted STARR-seq), that identifies allele-specific expression (ASE) while accounting for PCR duplicates through unique molecular identifiers. We tested 75,501 oligos (43,500 SNPs) and identified 2,720 SNPs with significant ASE (FDR 10%). To validate disruption of binding as one of the mechanisms underlying ASE, we developed a new high throughput allele specific binding assay for NFKB-p50. We identified 2,951 SNPs with allele-specific binding (ASB) (FDR 10%); 173 of these SNPs also had ASE (OR=1.97, p-value=0.0006). Of variants associated with complex traits, 1,531 resulted in ASE and 1,662 showed ASB. For example, we characterized that the Crohn's disease risk variant for rs3810936 increases NFKB binding and results in altered gene expression.

A general framework for predicting the transcriptomic consequences of non-coding variation

“…To assess peaBrain’s performance in predicting individual variation in RNA expression levels in comparison to other widely-used in silico methods and experimental assays (elastic net 6 , DeepSEA 8 , MPRA 11 , BiT-STARR-seq 12 , and HiDRA 13 ), we designed four tasks ( tasks E-H ; described below and summarized in Supplementary Table 2 ). For the comparison with elastic net, in line with other recent studies in the field 9 , we restricted performance analyses to a set of genes with significant non-zero narrow-sense cis -heritability (henceforth, simply referred to as heritability) in LCLs as estimated by constrained GCTA 30 (limited to the 1Mbps input sequence; p<0.01; see Online Methods ).…”

“…The sequence annotations were derived from the Roadmap’s GM12878 lymphoblastoid cell line 15-state ChromHMM model; the same GM12878 cell line was also used for both experimental assays (MPRA and HiDRA). BiT-STARR-seq was also performed in a lymphoblastoid cell line, but the exact cell line was not specified 12 . For each chromatin annotation, we assessed significance using a simple logistic model after rank-transformation of all estimates to normality (to ensure coefficients were comparable; see Online Methods ).…”

“…We further show that CNNs outperform linear models in predicting the consequences of genotype variation (Stage 2). In EBV-transformed lymphocytes (LCLs), we demonstrate that the estimated peaBrain variant effects correlate more strongly with coefficients from the univariate eQTL analysis, compared to log-skew effect estimates obtained from massively parallel reporter assays (MPRAs) 11 and bi-allelic targeted STARR-seq (BiT-STARR-seq) 12 , or log fold changes (logFC) of perturbed epigenetic states from DeepSEA 8 . To highlight the utility of the Stage 2 models, we identified putatively functional eQTLs in LCLs that are missed by experimental high-throughput approaches that characterise variant function, such as MPRAs, BiT-STARR-seq, and high-definition reporter assays (HiDRA) 13 .…”

Characterization of caffeine response regulatory variants in vascular endothelial cells

“…Massively parallel reporter assays (MPRA) have allowed studies of non-coding genetic variants and their role in gene regulation, at unprecedented scale (Arnold et al, 2013; Gordon et al, 2020; Kalita et al, 2018; Melnikov et al, 2012; Patwardhan et al, 2012; Tewhey et al, 2016; Ulirsch et al, 2016; Vockley et al, 2015; Wang et al, 2018). Originally developed to study the gene regulatory potential of promoters and enhancer sequences, MPRA protocols have been further developed to study regulatory genetic variation and fine map association signals (Kalita et al, 2018; Tewhey et al, 2016; Ulirsch et al, 2016; Vockley et al, 2015). MPRAs with synthetic regulatory sequences can test allelic activity for candidate regulatory variants independently of their allele frequency in the population (Kalita et al, 2018; Tewhey et al, 2016; Ulirsch et al, 2016).…”

“…Originally developed to study the gene regulatory potential of promoters and enhancer sequences, MPRA protocols have been further developed to study regulatory genetic variation and fine map association signals (Kalita et al, 2018; Tewhey et al, 2016; Ulirsch et al, 2016; Vockley et al, 2015). MPRAs with synthetic regulatory sequences can test allelic activity for candidate regulatory variants independently of their allele frequency in the population (Kalita et al, 2018; Tewhey et al, 2016; Ulirsch et al, 2016). .…”