Fig. 6: Inter-species translation of serology data.
From: AutoTransOP: translating omics signatures without orthologue requirements using deep learning

a Per feature Pearson correlation in ten fold cross-validation for human features and for non-human primate (NHP) features. b Classifiers’ performance in various tasks (n = 10 per task). In all boxplots, the centerline denotes the median, the bounds of the box denote the 1st and 3rd quantiles, and the whiskers denote points not being further from the median than 1.5 × interquartile range (IQR). c Functional grouping of NHP features predictive of protection-associated human features. In the top nightingale rose plot, NHP features are categorized by antigenic target. Blue features are antigens found on the surface of the HIV virus, green features are internal viral components, and orange features are specific to the antigens present in the primary vaccine (Prime) and booster vaccine (Boost). In the bottom nightingale rose plot, NHP features are categorized by serological feature type. Brown features are binding affinities specific to the human Fc-receptors, blue features are binding affinities specific to the NHP Fc-receptors, and yellow features are either anti-Rhesus IgG titers or functional assays (IFNg Elispots; C1q complement assays). In both plots, median percentile rank is plotted for all features belonging to each category. d Network visualization of the associations between specific NHP and human serological features, related to human protection. Brown nodes represent NHP features and dark blue nodes represent human features. Red and blue edges are connected to NHP features that are positively or negatively associated with protection in humans, respectively. Only the strongest NHP-human feature pairs are visualized here (see the methods for selection criteria).