ABC-GWAS: Functional Annotation of Estrogen Receptor-Positive Breast Cancer Genetic Variants

Abstract

Over the past decade, hundreds of genome-wide association studies (GWAS) have implicated genetic variants in various diseases, including cancer. However, only a few of these variants have been functionally characterized to date, mainly because the majority of the variants reside in non-coding regions of the human genome with unknown function. A comprehensive functional annotation of the candidate variants is thus necessary to fill the gap between the correlative findings of GWAS and the development of therapeutic strategies. By integrating large-scale multi-omics datasets such as the Cancer Genome Atlas (TCGA) and the Encyclopedia of DNA Elements (ENCODE), we performed multivariate linear regression analysis of expression quantitative trait loci, sequence permutation test of transcription factor binding perturbation, and modeling of three-dimensional chromatin interactions to analyze the potential molecular functions of 2,813 single nucleotide variants in 93 genomic loci associated with estrogen receptor-positive breast cancer. To facilitate rapid progress in functional genomics of breast cancer, we have created "Analysis of Breast Cancer GWAS" (ABC-GWAS), an interactive database of functional annotation of estrogen receptor-positive breast cancer GWAS variants. Our resource includes expression quantitative trait loci, long-range chromatin interaction predictions, and transcription factor binding motif analyses to prioritize putative target genes, causal variants, and transcription factors. An embedded genome browser also facilitates convenient visualization of the GWAS loci in genomic and epigenomic context. ABC-GWAS provides an interactive visual summary of comprehensive functional characterization of estrogen receptor-positive breast cancer variants. The web resource will be useful to both computational and experimental biologists who wish to generate and test their hypotheses regarding the genetic susceptibility, etiology, and carcinogenesis of breast cancer. ABC-GWAS can also be used as a user-friendly educational resource for teaching functional genomics. ABC-GWAS is available at http://education.knoweng.org/abc-gwas/.

Keywords: GWAS; breast cancer; functional characterization; variant annotation; web resource.

FIGURE 1

A flowchart showing integrative analysis pipeline used in ABC-GWAS. For each GWAS locus, we perform eQTL analysis, motif analysis, and gene-TF expression correlation analysis to obtain candidate (SNP, gene, TF) triplets. Blue boxes show analysis steps, orange boxes indicate data and/or resources used in the analysis, and text in red shows intermediate results.

FIGURE 2

A snapshot of the homepage of ABC-GWAS. Selecting a GWAS SNP using the left-hand-side drop-down menu populates the table on the right with relevant GWAS publications. Upon selecting an LD SNP or a CCV and clicking on the “Submit” button, various tabs on the bottom containing analysis modules are loaded.

FIGURE 3

Snapshots of some of the ABC-GWAS analysis modules. (A) Embedded genome browser showing the queried LD SNP location (vertical line) and ChIP-seq tracks in MCF-7 and T-47D cell lines. The “SNPs” track shows the locations of the GWAS SNPs and high LD SNPs, while the “CCVs” track shows the locations of the credible causal variants and their GWAS lead SNPs. (B) The plot shows average predicted chromatin contact counts across several HiC-Reg models in MCF-7 as a function of genomic location centered at the queried LD SNP (vertical line). The predictions for which at least one model shows significance (q < 0.05) is filled in red. The size of the markers is proportional to the number of models showing significance. (C) Normalized DNA copy number for normal (blue) and tumor (red) samples from TCGA in a 2 Mb window centered at the GWAS SNP location.

FIGURE 4

Case study of GWAS locus at TOX3: (rs4784227, TOX3, FOXA1). (A) Embedded genome browser showing the SNP rs4784227 within an open chromatin region (MCF-7 DNase tracks) and several TF peaks (“ReMap 2018 Peaks”). (B) eQTL plot showing significant correlation between TOX3 expression and rs4784227 genotypes. (C) Predicted TF candidate FOXJ3 motif from Kheradpour and Kellis (2014).

FIGURE 5

Case study of GWAS locus at ZMIZ1: (rs1250003, ZMIZ1, GATA). (A) Embedded genome browser showing the LD SNP rs1250003 within an open chromatin region (MCF-7 DNase track) and several TF peaks (“ReMap 2018 Peaks”). Black arrow indicates a chromatin looping interaction between the SNP locus and the ZMIZ1 promoter. (B) eQTL results showing the top significant target gene of the GWAS SNP to be ZMIZ1. (C) Average predicted chromatin contact counts across several HiC-Reg models in MCF-7, showing the ZMIZ1 promoter as one of the loci significantly interacting with the LD SNP. (D) Predicted TF candidate GATA motif from Kheradpour and Kellis (2014).