USING CHROMATIN CONFORMATION DATA TO INFER THE FUNCTIONAL IMPACT OF DISEASE-ASSOCIATED (SNPs)

Background Thousands of SNP-trait associations have been reported in published GWAS. However, taking the next step to identify the functional impact of these SNPs remains a significant challenge, due in part to the fact that many associated SNPs are located in non-coding regions of the genome. We have developed a new framework for annotating intergenic SNPs to target genes; we do this by integrating data about the 3D structure of chromosomes with information about predicted regulatory regions. This framework builds upon evidence that (a) chromatin looping events facilitate long-range interactions between enhancers and promoters, and (b) these interactions are important for regulating gene expression. We used results from two GWAS (ADHD and schizophrenia) conducted by the Psychiatric Genomics Consortium to investigate whether this novel annotation method can reveal new candidate genes targeted by regulatory variants. Methods SNPs were assigned to genes if they met any of the following criteria: 1) the SNP was located strictly within the boundaries of the gene, 2) the SNP was located within a promoter region that overlaps the TSS of any of the gene's transcripts, 3) the SNP was located within an enhancer that is brought into close proximity of the gene's promoter through chromatin looping. We required that both regulatory elements (promoters and enhancers) be within 100 kb of the loop anchors (the region of direct chromatin contact). Publicly-available high-resolution chromatin contact data, and promoter and enhancer regions from the ENCODE Project were used. GWAS summary statistics were retrieved from large-scale meta-analyses conducted by the Psychiatric Genomics Consortium. The schizophrenia meta-analysis included 36,989 cases and 113,075 controls. The ADHD meta-analysis included 20,183 cases and 35,191 controls. Results For the ADHD and schizophrenia analyses, respectively, 158,650 (59 with p ≤ 1e-5) and 187,867 (1005 with p ≤ 1e-5) SNPs were located within predicted enhancer regions. A substantial proportion (N=45,247, or 29% for ADHD; N=53,233, or 28% for schizophrenia) of these enhancer SNPs could be mapped to target genes through chromatin looping events. For ADHD, 18 SNPs with a p-value ≤ 1e-5 were mapped to 19 target genes through chromatin loops. Of these 19 genes, 11 are farther than 100 kb from the associated SNP, making it unlikely that they would be identified by conventional methods. Similarly, for schizophrenia, 85 target genes farther than 100 kb away from an associated enhancer SNP (p ≤ 1e-5) were identified through chromatin loop events. Using a manually curated PPI network of high-confidence interactions, subnetworks containing candidate genes and all common interactors (gene's connecting two candidate genes) were extracted. For both disorders, subnetworks that included genes implicated by chromatin looping events showed increased mean connectivity compared to subnetworks containing only genes near (within 1 kb) an associated SNP. Discussion Our results show that integrating data on the 3D structure of the genome with information about genomic regulatory elements has the potential to greatly improve the knowledge gained from GWAS and to generate new hypotheses about the genetic mechanisms of susceptibility to complex diseases.