Figure 2

The meta-analysis results show improved concordance with previous animal proteomic and human GWAS findings. (a) Venn diagrams comparing differentially-expressed genes from each meta-analysis (pfp < 0.05) with differentially-expressed genes from the individual studies that were meta-analyzed (as per the original authors’ statistical criteria). Note that most of the genes identified in the rank product meta-analyses were unique (i.e., not classified as differentially-expressed in the original transcriptome studies). (b) Heat map comparing the overlap of original transcriptome study results and rank product meta-analysis results with the genes implicated in previous GWAS and proteomics studies of refractive error. The shared category includes genes implicated in both the rank product meta-analyses and the individual studies included in the meta-analyses. Meta-analysis and original study categories include the remaining genes that were uniquely implicated in either the rank product meta-analysis or the original studies. The number of genes implicated in each list is shown in parenthesis after the list name. List intersections (i.e., the number of overlapping genes), Benjamini–Hochberg adjusted p-values, and odds ratios are superimposed on the heat map grid. Statistically significant overlaps are highlighted in red. Note that, although there were generally more commonalities in total between the original transcriptome study findings and the proteomics/GWAS results, the list of genes implicated in the original studies was very large resulting in low statistical significance and odds ratios. Supplementary Table S3 provides further details on all of the overlapping genes. (c) Venn diagram showing the rank product meta-analysis genes that were also implicated in previous GWAS or proteomics studies (i.e., all overlapping genes from the right two columns of Fig. 2b). Shared findings are shown in bold italic font.