Fig. 4.

The comparison of prediction accuracy performances of LMM and KAML. The performance was measured by the AUROC and Pearson correlation for the datasets of human and other species, respectively. Half represents that KAML randomly selected only half of the total individuals to optimize the model parameters; Adaptive represents running KAML on the entire dataset, and the parameters were optimized for each replicate. For each boxplot, the middle line represents the average value, the bottom and top are the standard deviation, and the upper and lower ends of each box represent the maximum and minimum, respectively