Fig. 1From: Universal clinical Parkinson’s disease axes identify a major influence of neuroinflammationIdentifying the underlying phenotypic axes using a Bayesian Mixed Model that incorporates genetic similarity. a Schematic representation of the approach used to capture the latent axes of clinical variation in deeply phenotyped cohorts. The method (PHENIX) was initially developed to impute missing observations (e.g. phenotype A) according to other available observations (here B). This approach also exploits genetic relationships where phenotypic heritability can be used to increase imputation accuracy. Here, we identify the relationship (diagonal blue) derived by PHENIX, herein named phenotypic axis, to capture the clinical variation. b Workflow of the analyses performed here. In each of the three cohorts, we independently derived the phenotypic axes associated with the baseline clinical variation and the phenotypic axes associated with the clinical progression. We also derived the phenotypic axes by exploiting the full genotype available or, instead by selecting the genotype in a subset of loci associated with a specific disorderBack to article page