{"doi":"10.7554/elife.10557.031","title":"Author response: Discovery and validation of sub-threshold genome-wide association study loci using epigenomic signatures","abstract":"Article Figures and data Abstract eLife digest Introduction Results Discussion Materials and methods Data availability References Decision letter Author response Article and author information Metrics Abstract Genetic variants identified by genome-wide association studies explain only a modest proportion of heritability, suggesting that meaningful associations lie 'hidden' below current thresholds. Here, we integrate information from association studies with epigenomic maps to demonstrate that enhancers significantly overlap known loci associated with the cardiac QT interval and QRS duration. We apply functional criteria to identify loci associated with QT interval that do not meet genome-wide significance and are missed by existing studies. We demonstrate that these 'sub-threshold' signals represent novel loci, and that epigenomic maps are effective at discriminating true biological signals from noise. We experimentally validate the molecular, gene-regulatory, cellular and organismal phenotypes of these sub-threshold loci, demonstrating that most sub-threshold loci have regulatory consequences and that genetic perturbation of nearby genes causes cardiac phenotypes in mouse. Our work provides a general approach for improving the detection of novel loci associated with complex human traits. https://doi.org/10.7554/eLife.10557.001 eLife digest Most complex traits are governed by a large number of genetic contributors, each playing only a modest effect. This makes it difficult to identify the genetic variants that increase disease risk, hindering the discovery of new drug targets and the development of new therapeutics. To overcome this limitation in discovery power, the field of human genetics has traditionally sought increasingly large groups, or cohorts, of afflicted and non-afflicted individuals. Studies of large cohorts are a powerful approach for discovering new disease genes, but such groups are often impractical and sometimes impossible to obtain. However, it has become possible to complement the genetic evidence found in disease association studies with biological evidence of the effects of disease-associated genetic variants. Wang et al. focus specifically on genetic sites, or loci, that do not affect protein sequence but instead affect the non-coding control regions. These are known as enhancer elements, as they can enhance the expression of nearby genes. These loci constitute the majority of disease regions, and thus are extremely important, but their discovery has been hindered by our relatively poor understanding of the human genome. Chemical modifications known as epigenomic marks are indicative of enhancer regions. By studying the factors that affect heart rhythm, Wang et al. show that specific combinations of epigenomic marks are enriched in known trait-associated regions. This knowledge was then used to prioritize the further investigation of genetic regions that genome-wide association studies had only weakly linked to heart rhythm alterations. Wang et al. directly confirmed that genetic differences in “sub-threshold” regions indeed alter the activity of these regulatory regions in human heart cells. Furthermore, mutating or perturbing the predicted target genes of the sub-threshold enhancers caused heart defects in mouse and zebrafish. Wang et al. have demonstrated that epigenome maps can help to distinguish which sub-threshold regions from genome-wide association studies are more likely to contribute to a disease. This allows for the discovery of new disease genes with much smaller cohorts than would be needed otherwise, thus speeding up the development of new therapeutics by many years. https://doi.org/10.7554/eLife.10557.002 Introduction Genome-wide association studies (GWAS) hold the promise of identifying genetic loci that drive complex disease, however realizing this goal has been challenging due to the modest effect sizes of most common variants that require extremely large cohorts to detect with significance. The recent demonstration that disease-associated single nucleotide polymorphisms (SNPs) reside preferentially in enhancer elements provides a unique opportunity to leverage epigenomic maps of regulatory elements for understanding the function of known GWAS loci and for prioritizing new loci missed in current studies (Ernst et al., 2011; Cowper-Sal·lari et al., 2012; Maurano et al., 2012; Trynka et al., 2013). Despite increasingly large GWAS cohort sizes, the current catalog of genome-wide significant loci still explains only a modest proportion of the heritability for any given trait, with an excess of low p-value loci still below the genome-wide significance threshold (Arking et al., 2014). These observations suggest that many more signals with 'sub-threshold' significance remain to be identified, however, the recognition of biologically relevant sub-threshold loci is hindered by a higher false positive rate (Hindorff et al., 2009; Maher, 2008; Altshuler et al., 2008). Thus, new computational approaches that integrate genetic data with genome-wide epigenomic profiles are needed to use existing cohorts to discover new loci and genes that influence complex traits and diseases. Here, we use epigenomic maps of 127 tissues from the Roadmap Epigenomics Project as a guide to systematically identify biologically relevant sub-threshold variants (Roadmap Epigenomics Consortium, 2015). As proof of concept, we focused on two cardiac traits with clinical significance: electrocardiographic QT interval reflecting myocardial repolarization and QRS duration reflecting cardiac conduction. These two traits have a clear tissue of origin and published GWASs have reported over a hundred QT/QRS loci, making these traits ideal for testing variants with sub-threshold significance (Supplementary file 1) (Hindorff et al., 2009; Maher, 2008; Altshuler et al., 2008). In particular, variation within QT interval length plays an important role in human disease, where extreme QT prolongation is associated with sudden cardiac death and can occur as an unintended side effect of many non-cardiac medications (Rabkin et al., 1982; Heist and Ruskin, 2010). We combine genome-wide maps of cardiac enhancer activity with the results from a large study of QT interval duration to identify dozens of novel QT loci with sub-threshold statistical significance. We provide multiple lines of evidence to show that these sub-threshold loci can alter enhancer activity, and we implicate specific genes through which these loci act to influence QT interval length. Importantly, we demonstrate that epigenetic signals can distinguish true biological signals from noise, thus bypassing the higher false positive rate that has previously hindered study of sub-threshold loci. We expect our work will uncover new genes involved in cardiac electrophysiology, aid in the identification of patients at risk for sudden cardiac death, and enable development of new treatments for susceptible individuals. More broadly, our work demonstrates the power of integrating epigenomics with existing GWAS to discover sub-threshold genetic loci and novel genes associated with complex human disease. Results QT/QRS-associated variants are enriched in cardiac enhancers We compiled a list of 112 QT/QRS loci from the NHGRI GWAS database (accessed July 2013, Supplementary file 1) and identified SNPs in strong linkage disequilibrium (r2>0.8) using genotype data from the 1000 Genomes Project (Phase 1, CEU population) (1000 Genomes Project Consortium, 2010). We also collected GWAS loci from a later meta-analysis of QT interval studies, published in June 2014 by Arking et al., which we held out from the aforementioned 112 QT/QRS loci as a validation dataset for subsequent analyses (Arking et al., 2014). Because only 22 of 112 loci (20%) harbor SNPs that overlap exons, we examined whether QT/QRS variants are enriched in predicted enhancer elements across the genome using chromatin maps across 127 tissues generated by the Roadmap Epigenomics Project including adult left ventricle (LV), adult right ventricle (RV), fetal heart (FH) and adult right atrium (RA) (Roadmap Epigenomics Consortium, 2015). QT/QRS variants have greatest overlap with predicted enhancers (as defined by high levels of H3K4me1 and low H3K4me3 using ChromHMM) from the four cardiac tissues compared to the other 123 non-cardiac tissues (red circles, Figure 1b, Supplementary file 1) (Ernst et al., 2011). This enrichment persists over randomly sampled sets of control loci with matched genetic properties such as minor allele frequency, number of SNPs in LD, distance to nearest gene, number of nearby genes, and presence on an Affymetrix 660W genotyping array (Figure 1a, Materials and methods). Enhancers from the LV showed the strongest enrichment of the four cardiac tissues (z-scores=7.67, empirical p<1x10−5, 105 permutations), demonstrating that an unbiased analysis can resolve the causal tissue with high precision, as QT interval and QRS duration are primarily reflective of myocardial repolarization in the ventricles. Figure 1 with 3 supplements see all Download asset Open asset GWAS repolarization loci preferentially overlap cardiac enhancers. (a) Enrichment of human left ventricle enhancers in 112 QT/QRS loci. The number of loci that contain a SNP overlapping an enhancer are computed for the 112 QT/QRS loci, and compared against 100,000 permutations of randomly sampled control loci matched for LD block size (number of SNPs), MAF, distance to nearest gene, number of nearby genes, and presence on genotyping array. (b) Top, Enrichment of enhancers from 127 human tissues in QT/QRS loci. Bottom, Enrichment of enhancers from non-cardiac tissues for QT/QRS loci is substantially weaker following removal of enhancers active in any of the four cardiac tissues. (c) Top, QT/QRS SNPs are more likely to disrupt motifs corresponding to expressed TFs compared to 100,000 sets of matched control loci. Bottom, Weaker enrichment was observed between repolarization and matched control loci when the sequence of the TF motif was randomly shuffled and re-mapped to the genome (10,000 permutations). https://doi.org/10.7554/eLife.10557.003 Focusing on the left ventricle, we analyzed the enrichment of both coding annotations using GENCODE and non-coding annotations using individual chromatin marks and chromatin states defined by ChromHMM as well as DNase I hypersensitivity (DHS) maps available in heart tissue (Ernst et al., 2011; Harrow et al., 2012; Thurman et al., 2012). We observed that intergenic enhancers are the most strongly enriched annotated genomic region (z-score > 7.5) in QT/QRS loci, followed by gene transcription regions (z-score between 3 and 6) (Figure 1—figure supplements 1 and 2). This enrichment increased significantly (z-score from 7.67 to 9.31 for left ventricle) when restricting the analysis to 'strong' enhancers (i.e. H3K4me1 enhancers that are also marked by H3K27ac). Our results indicate that predicted enhancers are highly informative for annotating trait-associated variants compared to other classes of genomic regions. Figure 2 Download asset Open asset Enhancers overlapping QT/QRS loci differ in functional characteristics from all enhancers. Several functional characteristics were compared between enhancers overlapping QT/QRS loci (red) and non-GWAS left ventricle enhancers (blue). Fold change represents fold change between median values for the two groups, and p-values were calculated using the Mann-Whitney U test. See Materials and methods for comparison methodology between GWAS QT/QRS enhancers and non-GWAS enhancers for each functional or epigenomic feature. For primate conservation, LV enhancers (blue) were size-matched (+/-1 kb) to GWAS enhancers to control for skewed enrichments driven by larger GWAS enhancer size. https://doi.org/10.7554/eLife.10557.007 We next asked whether LV enhancers that overlap QT/QRS loci have features that distinguish them from putative LV enhancers identified by ChromHMM that do not overlap QT/QRS loci (Figure 2). First, we considered the density of H3K27ac marks, as the co-enrichment of H3K4me1 and H3K27ac correlates with strong enhancer activity (Creyghton et al., 2010; Rada-Iglesias et al., 2011). We found that the 65 enhancers overlapping 45 QT/QRS loci have a 3.1-fold higher H3K27ac density compared to non-GWAS LV enhancers (p=1.53x10--4). In fact, incorporating H3K27ac into ChromHMM enhancer predictions resulted in substantially greater enrichment of QT/QRS loci (z-score = 9.31 vs. 7.67 for left ventricle); 44 of the 45 QT/QRS loci overlap an H3K27ac-defined 'strong' enhancer. QT/QRS LV enhancers are also more likely to be marked by either H3K4me1 or H3K27ac in at least one of the other three heart tissues (fetal, right atrium, right ventricle) compared to non-GWAS LV enhancers (p-values between 0.004 and 0.04, Figure 2) and less likely to be active in non-cardiac tissues (p=9x10-3, Figure 2). Left ventricular QT/QRS enhancers are significantly more hypomethylated than predicted LV enhancers not overlapping QT/QRS loci (hypomethylation p=1.07x10-6, hypermethylation p=0.60, Figure 2). Similar to H3K27ac, CpG hypomethylation correlates with increased enhancer activity, possibly through modulation of TF binding site accessibility (Hon et al., 2013; Stadler et al., 2011). Consistent with this idea, 22 of the 45 GWAS loci contain an enhancer SNP that alters a predicted motif for a cardiac-expressed TF (empirical p=0.002, 105 permutations) (Figure 1c). Moreover, QT/QRS GWAS enhancers are enriched for DHS and Cap Analysis Gene Expression (CAGE) signals in human fetal heart, both of which are marks of greater enhancer activity (Figure 2) (Thurman et al., 2012; Andersson et al., 2014). Finally, QT/QRS left ventricular enhancers show significant evolutionary conservation across the primate lineage compared to non-GWAS LV enhancers (p=6.82x10-5 compared to 105 size-matched sets of LV enhancers), suggesting that perturbation of these enhancers is under stronger negative selection. Taken together, QT/QRS loci preferentially overlap conserved enhancers that show cardiac-restricted activity, suggesting that common variants associated with these loci play roles in regulating cardiac functions that drive human phenotypes. Common features in GWAS cardiac enhancers identify novel sub-threshold loci Current GWAS loci collectively explain only a small fraction of the estimated heritability of a complex trait in part due to strict Bonferroni thresholds for multiple hypothesis testing (p<5x10-8) and the limited statistical power of existing studies to discover variants with modest effect sizes (Maher, 2008; Yang et al., 2011). We hypothesized that knowledge of the genomic properties associated with existing GWAS loci can guide the search for additional genetic signals that cannot be detected without increasing GWAS cohort sizes, and that these loci with weaker 'sub-threshold' p-values (i.e. 0.05>p>5x10-8) might reveal novel genes and biological pathways that contribute to complex disease. To test this idea, we used SNP summary statistics from the Arking et al. (2014) QT interval GWAS study we had earlier held out as a validation dataset (Arking et al., 2014). These summary statistics include the 112 QT/QRS loci identified by prior GWASs (red dots, bottom, Figure 3), as well as loci that reach genome-wide significance in the larger meta-analysis cohort but were not discovered in any previous GWAS (and therefore were not included in the 112 QT/QRS loci used for enrichment analyses above, gold dots, bottom, Figure 3). We observed that active LV enhancers are strongly enriched for loci harboring SNPs with p-values between 1x10-4 and 5x10-8 (Figure 3a, black line). Furthermore, the combination of functional features identified for above-threshold QT/QRS enhancers (Figure 2) substantially improves sub-threshold locus enrichment across a wide range of p-value thresholds (Figure 3a, colored lines, Figure 3—figure supplement 1). Figure 3 with 4 supplements see all Download asset Open asset Cardiac enhancers harbor additional sub-threshold QT loci. (a) Top, Enhancer characteristics learned on above-threshold QT/QRS loci from Figure 2 are predictive for additional sub-threshold loci (colored lines). Each point on a curve represents the fold difference in proportion of SNPs with p-value below the cutoff in the enhancer set versus the whole genome. Bottom, Manhattan plot of p-values for all SNPs from Arking et al. (2014) QT interval GWAS. 112 QT/QRS loci and all SNPs within 1 Mb are highlighted in red. Genome-wide significant loci newly discovered by Arking et al. and not in the 112 QT/QRS loci are highlighted in yellow. (b) Top, Enrichment signals for sub-threshold SNPs in left ventricle enhancers persists following removal of the 112 GWAS loci and nearby SNPs (+/- 1 Mb). Bottom, Manhattan plot of p-values for all SNPs from Arking et al. (2014) QT interval GWAS following removal of 112 QT/QRS loci and all SNPs within 1 Mb. Genome-wide significant loci newly discovered by Arking et al. and not in the 112 QT/QRS loci are highlighted in yellow. https://doi.org/10.7554/eLife.10557.008 Whether the enrichment of SNPs in the sub-threshold significance range represents linkage disequilibrium with existing above-threshold GWAS SNPs or novel biologically relevant loci remains an unresolved question (Maurano et al., 2012). In fact, an enrichment analysis using only SNPs nearby above-threshold GWAS loci produced a strong enrichment signature in the sub-threshold significance range (Figure 3—figure supplement 2). To distinguish between the two possibilities, we took a conservative approach and removed all SNPs within 1Mb of the initial 112 QT/QRS loci. Remarkably, the enrichment for LV enhancers persists and increases in the sub-threshold range (i.e. p=1x10-4 to 5x10-8, Figure 3b), likely due to removal of nominally significant SNPs that are in LD with above-threshold QT/QRS loci and do not represent true association signals. The enrichment for active LV enhancers in sub-threshold loci is not driven by biases in MAF, LD block size, distance to nearest gene, number of nearby genes, or presence on a SNP genotyping array (Figure 3—figure supplement 3). In total, we identified 2075 SNPs with p<1x10-4 that are independent of the 112 published QT/QRS loci, of which 208 SNPs overlap LV enhancers (Supplementary file 2). Epigenomic prioritization identifies sub-threshold loci with molecular functions Because the enrichment of sub-threshold SNPs in cardiac enhancers suggests that epigenetic prioritization can be used as a starting point for more in-depth investigations of sub-threshold signals from GWAS, we sought to directly test the molecular hypothesis that these sub-threshold loci impact the transcriptional regulation of cardiac genes (Figure 4a). We grouped all 2075 sub-threshold SNPs using linkage disequilibrium data (minimum r2=0.2) to identify 287 independent sub-threshold loci in the genome (Materials and methods). We prioritized loci where a sub-threshold SNP overlapped an active LV enhancer and either (i) also overlapped a fetal heart DNase I hypersensitivity peak or (ii) was an expression quantitative trait locus (eQTL) for a nearby gene. In total, we cloned allele-specific enhancer fragments from 22 cardiac enhancers that overlap SNPs from 18 independent sub-threshold loci, and performed quantitative luciferase assays in human iPSC-derived cardiomyocytes to determine whether the sub-threshold SNP genotypes influence enhancer activity (Materials and methods). We observed that 13 of 18 sub-threshold loci (72.2%) contain an enhancer that drives luciferase activity in an allele-specific manner (Figure 4b,d, Figure 4—figure supplement 1). Moreover, we estimate that between 51.1%-89.8% (95% Bayesian confidence interval) of prioritized sub-threshold loci show allele-specific activity on transcription, suggesting that the majority of sub-threshold loci identified by epigenomic prioritization do in fact have an impact on transcriptional enhancer activity. Figure 4 with 3 supplements see all Download asset Open asset Sub-threshold loci prioritized by epigenomics alter enhancer activity. (a) Model detailing how sub-threshold SNPs overlapping enhancers can affect QT interval. Green text: methods used to test mechanistic step in model. (b) Summary of luciferase enhancer reporter experiments. Left, luciferase enhancer reporter construct. Right, number of loci tested in panel d that exhibits significant allelic activity (p<0.05 between two haplotypes). (c), Left, schematic of a 3-D enhancer-promoter chromatin interaction detectable by 4C-seq. Right, number of loci tested in panel d where an enhancer-promoter interaction is observed in human iPS-derived cardiomyocytes by 4C-seq. (d) Experimental evidence that sub-threshold SNPs alter enhancer activity and that sub-threshold enhancers interact with gene promoters. Fold below GWS column represents degree to which sub-threshold locus is below genome-wide significance (5x10-8); Luciferase reporter column colored green if significant allelic difference in activity (p<0.05, Figure 4—figure supplement 1); Enhancer-promoter interactions column colored green if there is a detectable enhancer-promoter interaction by 4C-seq (Figure 4—figure supplement 3). https://doi.org/10.7554/eLife.10557.013 We also performed chromosome conformation capture combined with high-throughput sequencing (4C-seq) to experimentally test whether predicted enhancers in sub-threshold loci can form contacts with promoters, and to identify potential target genes of sub-threshold enhancers. We used 4C-seq to test ten predicted enhancers from eight sub-threshold loci in human iPSC-derived cardiomyocytes (van de Werken et al., 2012). Eight enhancers in six loci formed enhancer-promoter interactions in the proximal 500 kb region (Figure 4c, Figure 4 — figure supplement 3). These analyses provides evidence that the novel QT loci enhancers have regulatory activity and that the sub-threshold SNPs identified in our analyses can alter the activity of cardiac enhancers. Epigenomic prioritization discriminates biologically relevant sub-threshold loci We next tested whether epigenomic prioritization can distinguish biologically relevant sub-threshold loci by comparing properties of sub-threshold loci that do or do not overlap cardiac enhancers. From the 287 independent sub-threshold loci in the genome, we selected two subsets to compare: 60 loci that contain sub-threshold SNPs directly overlapping predicted active LV enhancers, and as a negative control, 129 sub-threshold loci that do not contain any SNPs (r2>0.2) overlapping a cardiac enhancer. Evidence from genome-wide association studies We reasoned that if sub-threshold loci that overlap active cardiac enhancers represent true biological signals, they should have stronger GWAS association signals than the negative control set. We present multiple lines of evidence supporting this hypothesis below (Figure 5a): The 60 enhancer-overlapping sub-threshold loci have significantly stronger p-values than the 129 negative control loci,  the  of the  p=1x10-4 threshold for both sets   Figure   of the 60 enhancer-overlapping sub-threshold loci are  the loci that reach genome-wide significance in the larger held out meta-analysis cohort (and not included in the 112 QT/QRS loci used for enrichment analyses in Figure 1 and  compared to only 3 of the 129 sub-threshold loci that do not overlap enhancers      The 60 enhancer-overlapping sub-threshold loci are more likely to reach  significance  in a  GWAS study of QRS duration  Materials and methods for   between both     In the QRS duration GWAS, p-values are available for  of 60 enhancer-overlapping sub-threshold QT loci and  of 129 negative control sub-threshold loci.  of   enhancer-overlapping sub-threshold loci are nominally significant in the QRS GWAS, a rate  higher than the 129 negative control loci  of  loci,   Figure  suggesting that epigenetic prioritization is more likely to identify sub-threshold SNPs that  in subsequent  These analyses demonstrate that genome-wide maps of predicted enhancers can  the detection of true sub-threshold loci. Evidence from organismal phenotypes Our identification of a  set of sub-threshold loci  on epigenomic signals provides a unique opportunity to discover new genes that contribute to cardiac  traits. As enhancers can  genes up to 1Mb  it is difficult to identify targets using a  nearest gene approach  et al.,  To  this  we  a computational    that  gene targets  on  activity  between   across  human tissues (Materials and methods).  this  we identified   genes predicted to be  by the 60 enhancer-overlapping sub-threshold loci (Supplementary file 2).   of the  observed 4C-seq interactions were predicted by our computational  compared to 3 of  by the   approach of  the enhancer target to the nearest gene. We used the  of the    to test whether these  genes have roles in QT interval. To this  we  mouse phenotypes for   and genetic  of the  predicted gene targets of sub-threshold enhancers. We identified  of the  genes where mouse   were available with  phenotypes  et al., 2015). Genetic perturbation in  of the  genes resulted in  cardiac  or cardiac  both  that are also  by genes nearby above-threshold QT interval loci and genes  in the   QT  This represents a  enrichment compared to genes linked to all active LV enhancers  of   black  Figure  In  phenotypes  from genetic perturbation of  genes nearby the 129 negative control sub-threshold loci  enhancers are  less likely to  in  cardiac  or  compared to our 60 prioritized sub-threshold loci  perturbation of 2 of 65 genes nearby the negative control  have relevant cardiac     Figure  The study of biologically relevant sub-threshold loci has been  by a high false positive rate that makes the  investigation of any sub-threshold locus experimentally more difficult and less  than above-threshold loci. The data   provide multiple independent lines of evidence that epigenomic  can be used to prioritize sub-threshold GWAS loci with a significantly greater  of  biologically  Sub-threshold locus at    enhancer activity  a  small number of above-threshold GWAS loci, including  for   the  locus for  and the  locus for QRS  have been  in   et al., 2010;    et al., 2012;    et al.,   et al., 2012;  et al., 2014). These studies all identified SNPs within non-coding regulatory elements that disrupt expression of a nearby gene that plays a  role in  a human  In   sub-threshold locus has been experimentally  or  to  We selected one locus on chromosome  where our results from Figures 4 and  suggest that sub-threshold SNPs disrupt enhancer activity and therefore expression of a gene involved in cardiac  We set out to  whether this locus can  as an  for  investigations of other sub-threshold loci. Figure  Download asset Open asset Epigenomic prioritization  biologically relevant sub-threshold loci (a) Left, Sub-threshold loci overlapping enhancers have significantly stronger association signals than loci  enhancers in the QT interval GWAS.   overlapping enhancers have significantly more likely to be newly genome-wide significant in the  QT interval GWAS, than loci  enhancers. Right, Sub-threshold loci overlapping enhancers are significantly more likely to be nominally significant  in QRS GWAS than sub-threshold loci not overlapping enhancers. (b) Genetic perturbation of genes with predicted  to 60 enhancer-overlapping sub-threshold loci are significantly more likely to  cardiac  or  phenotypes than genes linked to all LV enhancers and genes nearby  overlapping sub-threshold loci.  The sub-threshold locus on chromosome    SNPs with reported p-values less than 1x10-4 and  2 SNPs in LD that do not have calculated  We focused on the 3 SNPs in this locus that overlap active LV    and   not  that both overlap a  kb predicted  and   that  a","journal":null,"year":2016,"id":10230,"datarank":0.0,"base_score":0.0,"endowment":0.0,"self_citation_contribution":0.0,"citation_network_contribution":0.0,"self_endowment_contribution":0.0,"citer_contribution":0.0,"corpus_percentile":null,"corpus_rank":null,"citation_count":0,"citer_count":0,"citers_with_citation_signal":0,"citers_with_endowment":0,"datacite_reuse_total":0,"is_dataset":false,"is_dataset_confidence":0.074,"is_data_producer":false,"deposit_databanks":null,"is_oa":true,"file_count":0,"downloads":0,"has_version_chain":false,"published_date":"2016-03-23","fair_score":null,"fair_percentile":null,"algorithm_id":"datarank_citation_only_1hop_v6","ranking_scope":"data_only","authors":[{"id":53932,"name":"Nathan R. Tucker","orcid":"0000-0002-5071-4218","position":1,"is_corresponding":false},{"id":53933,"name":"Gizem Rizki","orcid":null,"position":2,"is_corresponding":false},{"id":53934,"name":"Robert Mills","orcid":null,"position":3,"is_corresponding":false},{"id":53935,"name":"Peter H.L. Krijger","orcid":"0000-0003-1702-348X","position":4,"is_corresponding":false},{"id":53936,"name":"Elzo de Wit","orcid":"0000-0003-2883-1415","position":5,"is_corresponding":false},{"id":42114,"name":"Vidya Subramanian","orcid":null,"position":6,"is_corresponding":false},{"id":53937,"name":"Eric Bartell","orcid":"0000-0002-4380-1401","position":7,"is_corresponding":false},{"id":53938,"name":"Xinh-Xinh Nguyen","orcid":null,"position":8,"is_corresponding":false},{"id":53939,"name":"Jiangchuan Ye","orcid":null,"position":9,"is_corresponding":false},{"id":53940,"name":"Jordan Leyton-Mange","orcid":null,"position":10,"is_corresponding":false},{"id":53941,"name":"Elena V Dolmatova","orcid":null,"position":11,"is_corresponding":false},{"id":21986,"name":"Pim van der Harst","orcid":"0000-0002-2713-686X","position":12,"is_corresponding":false},{"id":53942,"name":"Wouter de Laat","orcid":null,"position":13,"is_corresponding":false},{"id":896,"name":"Patrick T. Ellinor","orcid":"0000-0002-2067-0533","position":14,"is_corresponding":false},{"id":53943,"name":"Christopher Newton-Cheh","orcid":null,"position":15,"is_corresponding":false},{"id":53944,"name":"David J. Milan","orcid":"0000-0003-0149-3110","position":16,"is_corresponding":false},{"id":14693,"name":"Sharon L. R. Kardia","orcid":"0000-0002-9853-3379","position":17,"is_corresponding":false},{"id":53945,"name":"Laurie A. Boyer","orcid":"0000-0003-3491-4962","position":18,"is_corresponding":false},{"id":53946,"name":"Robert W. Mills","orcid":"0000-0003-4201-7137","position":19,"is_corresponding":false},{"id":53947,"name":"Jordan Leyton‐Mange","orcid":"0000-0001-8229-3326","position":20,"is_corresponding":false},{"id":53948,"name":"Elena Dolmatova","orcid":"0000-0001-9534-0742","position":21,"is_corresponding":false},{"id":53949,"name":"Christopher Newton‐Cheh","orcid":"0000-0002-3001-9111","position":22,"is_corresponding":false},{"id":53931,"name":"Xinchen Wang","orcid":"0000-0002-5198-6581","position":0,"is_corresponding":true}],"reference_count":0,"raw_metadata":null,"created_at":"2026-03-01T18:20:47.508186Z","pmid":null,"pmcid":null,"fwci":null,"citation_percentile":null,"influential_citations":0,"oa_status":null,"license":null,"views":0,"total_file_size_bytes":0,"version_count":0,"fair_f":null,"fair_a":null,"fair_i":null,"fair_r":null,"fair_zscore":null,"fair_rationale":null,"fair_model":null,"fair_agent_version":null,"fair_fulltext_source":null,"fair_has_llm":null,"fair_computed_at":null,"clinical_trials":[],"software_tools":[],"db_accessions":[],"linked_datasets":[],"topics":[]}