Figure 5.

Colocalisation analysis of Graves' and Hashimoto's diseases. Regions are labelled by chromosome and likely candidate gene(s) and arranged so that the top seven regions showed marginally significant association with both GD and HT ($P<0.05$) and the bottom six with just GD in the published single SNP analysis [Cooper et al., 2012]. The left panel shows the estimate of the coefficient of proportionality, η, for the estimate of the ratio of effect sizes in HT compared to GD, and its 95% credible interval, calculated using either the PCs approach (PC), or BMA averaging over two SNP models (BMA.2) or three SNP models (BMA.3). The middle panel shows the the evidence against colocalisation using either the $-{log}_{10}\left(P\right)$ value (PCs) or the posterior predictive $-{log}_{10}\left(P\right)$ (BMA). The right panel summarises the evidence for association of the region with each disease using $-{log}_{10}\left(P\right)$ values for the association analysis of Graves' and Hashimoto's using the selected principal components for the PCs approach. The $-{log}_{10}$ scale has been truncated at 10 so that more extreme P values are displayed at $-{log}_{10}\left(p\right)=10$. For the PC approach, testing was based on the smallest subset of components that captured 90% of the genetic variance.