Skip to main content
. 2021 Feb 15;11:610348. doi: 10.3389/fcimb.2021.610348

Figure 1.

Figure 1

Difference in balanced accuracy (bACC) of XGB models trained and evaluated under random CV and genome distance-aware CV for all considered organism/compound pairs. Significance thresholds are the probability of obtaining bACC estimates as low or lower than the ones from genome distance-aware CV when sampling from a normal distribution fitted to 10 random CV replicates obtained with different random seeds.