Fig. 8

Performance of the hybrid predictor. For each possible hybrid predictor in this work, we computed its MCC. In the figure we show the frequency histogram of these values. With dashed lines we show the prediction of the different reference methods, estimated on the original set of 59,442 variants. We see that hybrid methods tend to outperform reference methods, although this depends on the latter. For example, PON-P2 alone is better than many of the hybrids