Fig. 1: Modelling the relationship between genetic liability to schizophrenia and acne vulgaris.

A The CAUSE modelling framework was utilised to estimate the effect of genetic liability to schizophrenia on acne vulgaris (left), and conversely, genetic liability to acne vulgaris on schizophrenia (right). In both instances, the change in expected logwise point density (ΔELPD) is plotted, with error bars denoting the standard error, between the two models in three configurations: null versus sharing, null versus causal, and sharing versus causal. Three different beta priors were implemented: q ~ Beta(1,10), q ~ Beta(1,2), and q ~ Beta(1,100). An asterisk denotes a significant difference in model fit (P < 0.05). B The results of the MR-clust approach are visualised which investigated the association of individual schizophrenia IVs with acne. This is a mixture model framework that seeks to identify clusters of IVs with similar causal estimates. The size of the point denotes the cluster inclusion probability, relating to the conditional probability of cluster membership. The null cluster, coloured pink, relates to IVs with null effect, whilst the black “junk cluster” includes variants that were not parsimoniously assigned to any cluster. The cluster of interest is denoted in grey, with a trend line indicative of the mean cluster effect. The leftmost panel utilised default parameters for cluster assignment, whilst the right panel only retained variants with an inclusion probability > 80%.