Fig. 4

Accuracy of Pangenie and KAGE as a function of the number of individuals included in the model on a small test dataset. Recall and precision (shown as F1 score) are computed for the predicted genotypes by the various methods on a small test dataset (genome size 5 million bp) with reads simulated from HG002, in order to test whether the accuracy of PanGenie and KAGE increases with a larger number of individuals in the model. The F1 score of Malva and “our naive approach” are shown as a single dot, since both of these only use the whole population genotype frequency when estimating genotype probabilities