Distribution of AUC scores among 21 species and the correlation between predictive accuracy and divergence time.

A) Variation in prediction accuracy from different training sets to the same testing set. Each boxplot displays the variation in AUC scores calculated from different training sets (in addition to the species on the X axis, the rest of the 20 species were used as the training set) relative to the same testing set (the species on the X axis). ‘+’ represents the outliers corresponding to the training sets that have significantly lower or higher predictive accuracy. B) Heatmap matrix of the influence of training sets on different species. Colors in each cell indicate AUC scores obtained when those species were used as the training set to predict the essential genes of the target 21 species including the speciesself. CJE shows the worst prediction performance when CJE was used as the training set. C) Correlation between AUC scores and divergence times. Stars refer to the AUC scores obtained from the organisms with the divergence time on the X axis, in which one species is used as a training set and the other was used as a testing set. Red stars refer to outliers that were discarded from the regression analysis. The regression line is indicated by a solid line, and error bars are indicated by dashes. The boxplot shows the variation in prediction per 500 millions of years. D) Correlation between PPV scores and divergence times.