Leveraging the shared and opposing genetic mechanisms in the heritable cardiomyopathies

Figures

• 

  Extended Data Figure 1:. Forest plot of lead SNPs from DCM and HCM GWAS loci Effect sizes and 95% confidence intervals for lead variants at each locus are shown based on association HCM GWAS (green, left) and DCM GWAS (purple, right). Each point represents the estimated effect size (β) from the corresponding study, with horizontal bars indicating the 95% confidence interval. Loci are annotated to candidate genes using integrative locus-to-gene mapping approaches, and the significance of associations is indicated per trait. (Methods) To enable direct visual comparison of directionality, effect sizes, and alleles, the data were aligned across traits. Specifically, for each overlapping locus, we aligned effect sizes such that HCM effects were always visualized on the left and DCM effects on the right. To achieve this, we compared the direction of effect between traits and, if the DCM effect size was smaller than the HCM effect size, we flipped the effect alleles and corresponding β estimates. This harmonization step ensured that the more positive effect consistently appeared on the left (HCM) and the more negative or opposing effect on the right (DCM), facilitating direct visual comparison. Variant…

• 

  Extended Data Figure 2:. Quantile-quantile plots of contributing studies to the meta-analyses and MTAG of DCM, HCM, and CC, and shared meta-analysis. Each panel shows the quantile-quantile plot for a given GWAS, MTAG, or shared meta-analysis in a given dataset. In each quantile-quantile plot, the x-axis represents the expected −log10 of the P-value of variants under the null hypothesis, while the y-axis represents the observed −log10 of the P-value in the association study. The dashed line shows the expected calibration under the null hypothesis. For each QQ-plot, we report: i) the inflation factor λ, computed as the observed median χ2 statistic over the expected under the null, using all plotted variants; ii) the LDSC-based inflation factor (λLDSC), which restricts to high-confidence, common variants from LD Score Regression reference panels; iii) the interceptLDSC, which quantifies residual inflation due to confounding (e.g., population structure, relatedness), estimated as the intercept from regressing χ2 statistics on LD scores.22 Variants are filtered to those that passed filters for inclusion into the overall analyses (Methods). P-values are derived from various logistic regression models; reported…

• 

  Extended Data Figure 3:. Case-case approaches identify novel loci and genes underlying the cardiomyopathy spectrum. The outer left and right parts of the figure show rotated Manhattan plots, with genomic positions on the y-axis and −log10 of the association P-values on the x-axis; the left Manhattan shows results from CC-GWAS, while the right shows results from an MTAG incorporating CC-GWAS with relevant cardiac MRI traits (CC-MTAG). The middle of the figure shows a heatmap for all independent loci (N=113) identified across published DCM GWAS or MTAG (Jurgens et al.), published HCM GWAS or MTAG (Tadros et al.), CC-GWAS and CC-MTAG; green indicates that a given locus was identified in a given group of GWAS, including the present case-case approaches (CC), the published DCM study (DCM), or the published HCM study (HCM); ‘G’ indicates that a given locus was identified in a GWAS approach, ‘M’ suggests that a given locus was identified in an MTAG approach, and ‘GM’ suggests that a given locus was identified in both GWAS and MTAG. Loci are annotated to highly prioritized genes using contemporary gene-prioritization methods (Methods). Notably, novel loci are highlighted in dark green, and novel genes are…

• 

  Extended Data Figure 4:. Forest plot of lead SNPs from CC GWAS loci Lead variants from each locus from CC GWAS are plotted using association results from HCM GWAS on the left (green) and DCM GWAS on the right (purple); dots and error bars represent effect sizes with 95% confidence intervals. Loci are annotated to genes using contemporary locus-to-gene mapping (Supplementary Table 6)

• 

  Extended Data Figure 5:. Cell type enrichment highlights a specific enrichment of cardiomyocytes in cardiomyopathy spectrum heritability. Results are derived from a cell-type-specific enrichment pipeline, modified from sc-linker39, that tests for enrichment of gene programs derived from single-nucleus RNA-seq of the the left ventricle38 (N=18 healthy donors). Panel a shows results in the form of a dot plot for curated cell types (Supplementary Note), while panel b shows results for cell states. In both panels, the x-axis represents different cell types. In contrast, the y-axis represents the −log10 of the enrichment P-value (enrichment based on the Tau statistic conditional on the baseline model). In both panels, the black dotted line represents the Bonferroni-corrected significance cutoff, while the gray dotted line represents P=0.05. In panel b, the cell states are colored by the respective cell type to which they belong. Only cardiomyocyte cell types and cell states are enriched for heritability in our CC-GWAS data.

• 

  Extended Data Figure 6:. Pathway enrichment analysis of prioritized genes from case-case studies (CC GWAS/CC MTAG) using g:Profiler Pathway enrichment analysis using g:Profiler on genes prioritized from CC GWAS and CC MTAG loci. We considered 68 independent loci for CC GWAS and 17 novel loci identified in either CC GWAS or CC MTAG that were not reported in DCM or HCM GWAS/MTAG. If a locus was annotated with multiple genes (e.g., through different prioritization methods), all nominated genes were retained for downstream enrichment analysis. (Supplementary Table 8,10) Gene prioritization was performed using two complementary approaches: (1) FLAMES/PoPS-based prioritization, which integrates diverse functional annotations and similarity scores, and (2) a nearest-gene approach, where the gene closest to the lead SNP was selected. Panel a includes 77 prioritized genes from 68 CC GWAS loci (FLAMES/PoPS-based; query length = 77), and panel b consists of 68 genes based on nearest gene annotation (query length = 68). Panel c represents the 20 genes from the 17 novel loci using FLAMES/PoPS prioritization (query length = 20), and panel d shows the corresponding nearest gene set (query length = 18). Enrichment was…

• 

  Extended Data Figure 7.. Phenome-wide association heatmap with statistically significant associations for top variants of novel loci from CC GWAS/CC MTAG Heatmap showing standardized effect sizes (β) for lead genome-wide significant variants from PheWAS using data from the Cardiovascular Disease Knowledge Portal, specifically showing associations with cardiovascular-related traits that reached the threshold of statistical significance (P< 5.00e−8). On the y-axis, each column corresponds to a lead variant, annotated with its locus ID (in the format Chromosome:Position (hg19 (GRCh37) genome build):Reference allele:Alternate allele), locus number (Supplementary Table 11, Supplementary figure 10), and gene. Each row represents a phenotype grouped by trait clusters (color-coded). Cells are colored by the direction and magnitude of the effect (red = positive, blue = negative), with white indicating no significant association. Only associations with non-missing β estimates are displayed. Note: AF, Atrial fibrillation; ApoA, Serum ApoA; ApoB, Serum ApoB; CHOL, Total cholesterol; CK, Creatine kinase; DBP, Diastolic blood pressure; HDL, HDL cholesterol; HEIGHT, Height; HYPERTENSION, Hypertension; LATEF, Left…

• 

  Extended Data Figure 8:. Druggability of prioritized genes a, d. DrugnomeAI-derived druggability scores (Tclin + Tier1 probability) for prioritized genes with high translational potential (Methods), stratified by the presence (a) or absence (d) of a known pharmacological agent in the Open Targets Platform. Orange bars denote genes identified exclusively through CC-GWAS or MTAG analyses. The x-axis reflects the Tclin + Tier1 probability, a composite score estimating the likelihood that a gene is targetable based on existing clinical and regulatory precedent. b, e. Single-nucleus RNA-sequencing expression profiles38 for the same gene sets. Dot size represents the percentage of nuclei expressing the gene in a given cell type, while color indicates scaled mean expression. Supplementary Table 23 and Supplementary Figure 10. c. Drug development landscape for prioritized genes with known pharmacological agents. The first column indicates whether at least one drug targeting the gene is approved, with a representative agent listed. Subsequent columns summarize clinical indications for all drugs targeting each gene, categorized into cardiac diseases, muscle disorders, vascular/blood disorders, and cancer…

• 

  Extended Data Figure 9:. Manhattan plots for two stages of shared effect meta-analysis Both panels are Manhattan plots where each dot represents a genetic variant, with genomic positions on the x-axis and −log10 of the association P-value on the y-axis. Panels a shows results for meta-analysis statistics using a regular fixed-effects inverse-variance-weighted approach (3 loci). Panel b shows results after a second-stage meta-analysis under a random-effects framework (1 locus) (Methods)

• 

  Figure 1:. Study overview. a. Two Manhattan plots recreated from summary statistics of previously published case-control GWAS for DCM15 (37 loci; on the left) and HCM16 (33 significant loci; on the right). The plots highlight 18 loci with annotated genes shared between the two studies. b. Assessment of the genetic correlation between DCM and HCM on global and regional levels. Left plane: global genetic correlation was assessed using bivariate LD score regression (LDSC)22, which highlighted a strong inverse genetic correlation between DCM and HCM. Right plane: regional genetic correlations were evaluated across genome-wide regions using LAVA23, identifying 14 regions with significant inverse genetic effects between DCM and HCM. c. To model DCM and HCM as opposites on a single genetic spectrum, a statistical case-case GWAS (CC-GWAS) approach was implemented24; CC-GWAS was further combined with GWAS data for cardiac MRI traits in an MTAG41. These CC-GWAS approaches identified dozens of novel loci for the cardiomyopathy spectrum. d. To model DCM and HCM as similar diseases, a shared-effect meta-analysis was applied, which identified one novel locus. e. Contemporary locus-to-gene mapping was…

• 

  Figure 2:. Global and regional genetic correlations between DCM and HCM. a is a heatmap of bivariate genetic correlations estimated from GWAS data, showing DCM (on the left), HCM (on the right), and relevant cardiac MRI traits in between, including negative global circumferential strain (-Ecc), left ventricular end-systolic volume (LVESVi), and left ventricular concentricity (LVconc). The color represents the level of genetic correlation, with red and blue representing positive and negative correlation, respectively. (Supplementary Table 3) Panel b is a dot plot showing effects of lead variants from DCM and HCM GWAS, with estimated effects on DCM on the x-axis and effects on HCM on the y-axis. Loci are colored depending on whether they reached genome-wide significance in DCM GWAS (purple), HCM GWAS (green), or both (red). Error bars represent standard errors. (Extended Data Figure 1; Supplementary Table 4) Panel c shows an overview of the steps used to analyze regional bivariate genetic correlations between DCM and HCM, using LAVA. Panel d is a Manhattan plot showing results from LAVA, where each genomic region is represented by a single dot with genomic regions on the x-axis and the −log10…

• 

  Figure 3:. Modeling DCM and HCM as opposites on a disease spectrum yields improved locus discovery. All panels are Manhattan plots where each dot represents a genetic variant, with genomic positions on the x-axis and −log10 of the association P-value on the y-axis. Panels a and b show results for published DCM GWAS (37 loci) and HCM GWAS (33 loci), respectively. Panel c shows results from a case-case GWAS in which DCM and HCM are modeled as opposites on a single genetic spectrum, yielding 67 significant loci. Panel d shows results for a shared-effect meta-analysis in which DCM and HCM are treated as similar diseases, yielding only one significant locus.