Integrative Genomic Analysis Predicts Causative Cis-Regulatory Mechanisms of the Breast Cancer–Associated Genetic Variant rs4415084

Abstract

Running title: Regulatory genomics of GWAS SNPsAbbreviations: GWAS -genome-wide association study; eQTL -expression quantitative trait loci; ASE -allele-specific expression; TF -transcription factor; LDlinkage disequilibrium; FPKM -fragments per kilobase of transcript per million mapped reads; LCASE -local chromosome allele-specific expression; DHS -DNase I hypersensitive sites; PWM -position weight matrix; TCGA -The Cancer Genome Atlas; ER+ -estrogen receptor positive; TAD -topologically associated domain; MAF -minor allele frequency; RPKM -reads per kilobase of transcript per million mapped reads; SNP -single nucleotide polymorphism; MAPQ -mapping quality;ChIP-seq -chromatin immunoprecipitation sequencing; ASB -allele-specific binding;ncRNA -non-coding RNA; TSS -transcription start site; DNase-seq -DNase I Major FindingsResearch. Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on January 19, 2018; DOI: 10.1158/0008-5472.CAN-17-3486 Cancer Research Mathematical Oncology 4We developed a computational framework for integrating GWAS results with heterogeneous cancer genomic data and tissue-specific epigenetic data to facilitate the discovery of causative variants functioning through long-distance gene regulation.Applied to a breast cancer susceptibility region in 5p12, our method provides strong support for a putative causative SNP that is predicted to modulate GATA3 binding and regulate the expression of MRPS30 and nearby lncRNAs. Quick Guide to Equations and AssumptionsSince the majority of GWAS variants lie in non-coding regions of the human genome where a direct link to gene function is not obvious, we searched for (causative SNP, TF, target gene) triplets under the model of gene regulation by enhancers, in which the SNP interferes with the binding affinity of a key transcription factor (TF). With this assumption, we built a regulation model for a breast cancer susceptibility locus harboring three GWAS SNPs in the 5p12 region. To infer candidate target genes, we first performed expression quantitative trait loci (eQTL) analysis by regressing gene expression levels against two co-variates: genotype status at a given GWAS SNP and copy number of the gene. For each pair of ∈ {GWAS SNPs in 5p12} and ∈ {genes in 5p12 TAD}, the eQTL model can be expressed as: DOI: 10.1158/0008-5472.CAN-17-3486 Cancer Research Mathematical Oncology 5 hypothesis testing was further applied using ≤ , where is the total number of genes tested in the TAD ( = 22, thus = 0.0023).Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on JanuaryTo identify cis-regulated target genes, we tested local chromosome allele-specific expression (LCASE) using exonic SNPs that were properly phased with the GWAS SNP . For each exonic SNP , we obtained a subset of patients who had heterozygous genotypes at both the GWAS SNP and the exonic SNP . For each patient ( ∈ {1, ...

Differential Allele-Specific Expression Uncovers Breast Cancer Genes Dysregulated By Cis Noncoding Mutations

“…34 Furthermore, even if we find that a gene is dysregulated and suggest observed cis mutations as potentially causative, there may be multiple detected cis mutations and additional analysis is necessary to detect which of those are causing the change in ASE. 60 Nevertheless, because of the large number of noncoding mutations in cancer genomes, it is vital to have techniques for filtering out those mutations that are unlikely to be playing a regulatory role; indeed, we have found that the vast majority of noncoding mutations near genes do not appear to alter their expression.…”

“…However, none of the individuals included in our dataset had a TERT promoter mutation or the previously described functional recurrent promoter mutations in breast cancer, such as the ones in the FOXA1 promoter. 15,17,20,60 This is not surprising due to the low levels of recurrence of even the most well studied noncoding somatic mutations. As a result, the overlap in predicted functional mutations by different methods is expected to be small, and further development of methods such as ours that are not based on recurrence is critical.…”

Mechanistic interpretation of non-coding variants helps discover transcriptional regulators of drug response

“…Our direct comparisons between results (TF-drug associations) obtained with and without use of Delta-MOP scores (Table 1) are among the first direct statistical findings of the value of TFBS-SNP impact prediction for mechanistic studies of phenotypic variation, especially on genome-wide scales. More 'localized' applications, such as prioritization of a small number of candidate SNPs, are already being reported in the field (5). We also note that the basic methodology of our work can serve as a practical way to assess the value of new approaches to SNP scoring and prioritization, since the results of this methodology are testable findings at the TF and gene level (their roles in phenotype, as illustrated in Table 2 and Figure 4): there is more literature evidence to compare against at these levels than there is evidence for SNP function and mechanism.…”

“…A common approach is to find polymorphisms/variants that are statistically correlated with phenotypic differences, as in genome-wide association studies (GWAS) (1), familybased association tests (2), and expression quantitative trait loci (eQTLs) (3,4) for phenotype-related genes. However, statistically identified variants may not be functionally related to phenotypes (5), due to a variety of factors including linkage disequilibrium (LD). This problem is particularly pronounced in the case of non-coding variants, which represent the vast majority of GWAS findings (6,7) and often function by influencing gene regulation.…”

“…For example, Zhang et al (5) adopted such a strategy: they combined a method for predicting changes in TF binding with multi-omics data to identify a SNP that impacts the binding strength of a TF called GATA3 to modulate breast cancer susceptibility.…”