Fig. 4: Ranking of standardized marketed chemicals grouped by chemical classes. | Nature Communications

Fig. 4: Ranking of standardized marketed chemicals grouped by chemical classes.

From: Uncertainty-aware machine learning to predict non-cancer human toxicity for the global chemicals market

Fig. 4

Chemical classes are based on the lowest classification level (subclass, class (C) or superclass (S)) available in the ClassyFire chemical taxonomy. a, b The scatterplot of predicted POD and associated uncertainty given by the 95% CI for all available chemical classes and further highlighting the top 25 chemical classes – ranked by their toxicity potency given by the median predicted POD – covering at least 30 marketed chemicals with c, d the number of chemicals in the top 25 chemical classes, e, f the distribution of predicted PODs in the top 25 chemical classes and g, h the distribution of associated prediction uncertainty given by the 95% CI in the top 25 chemical classes. All data points are colored by their respective superclass among the 15 most frequent ClassyFire superclasses across the set of marketed chemicals (see legend). Boxplots show the median (center) and interquartile range (boxes) with whiskers extending to the 2.5th and 97.5th percentiles with underlying sample sizes shown in subfigures (c, d). CI confidence interval, POD point of departure. Source data are provided as a Source Data file.

Back to article page