Meta-Analysis of Genome-Wide Association Studies and Network Analysis-Based Integration with Gene Expression Data Identify New Suggestive Loci and Unravel a Wnt-Centric Network Associated with Dupuytren's Disease

Abstract

Dupuytren´s disease, a fibromatosis of the connective tissue in the palm, is a common complex disease with a strong genetic component. Up to date nine genetic loci have been found to be associated with the disease. Six of these loci contain genes that code for Wnt signalling proteins. In spite of this striking first insight into the genetic factors in Dupuytren´s disease, much of the inherited risk in Dupuytren´s disease still needs to be discovered. The already identified loci jointly explain ~1% of the heritability in this disease. To further elucidate the genetic basis of Dupuytren´s disease, we performed a genome-wide meta-analysis combining three genome-wide association study (GWAS) data sets, comprising 1,580 cases and 4,480 controls. We corroborated all nine previously identified loci, six of these with genome-wide significance (p-value < 5x10-8). In addition, we identified 14 new suggestive loci (p-value < 10-5). Intriguingly, several of these new loci contain genes associated with Wnt signalling and therefore represent excellent candidates for replication. Next, we compared whole-transcriptome data between patient- and control-derived tissue samples and found the Wnt/β-catenin pathway to be the top deregulated pathway in patient samples. We then conducted network and pathway analyses in order to identify protein networks that are enriched for genes highlighted in the GWAS meta-analysis and expression data sets. We found further evidence that the Wnt signalling pathways in conjunction with other pathways may play a critical role in Dupuytren´s disease.

Fig 1. Results of the Genome-Wide Meta-Analysis in Dupuytren’s Disease.

P-values were obtained under an additive genetic model with adjustment for minor population stratification. Effect sizes were synthesized using a fixed-effects regression model, thereby hypothesizing that the studies differed only in their statistical power to detect the outcome of interest. Both Cochran’s Q statistic and I² index were used to evaluate statistical heterogeneity between studies. (A) Quantile-quantile (QQ) plot of observed phenotypic association p-values. The grey line represents concurrence of the expected and the observed p-values under a true null hypothesis of no association. Values above the line indicate a signal in the data. (B) Manhattan plot showing the p-values (−log₁₀) plotted against their respective positions on each chromosome with each dot representing a SNP. Green dots represent SNPs that surpass the genome-wide significance threshold of 5×10⁻⁸. Manhattan and QQ plots were generated based on those 4,350,741 SNPs included in the meta-analysis that did not show indication of heterogeneity between studies.