Fig. 8
From: Large-scale comparison of machine learning methods for profiling prediction of kinase inhibitors
Comparison of average AUC values of A AtomPairs-, B FP2-, C MACCS-, D Morgan-, E PharmacoPFP-, and F RDKitDes-, G Graph-based models using five ML and one DNN DL methods. The average AUC values of the test sets for various ML and DL algorithms are displayed as boxplot. Middle spheres represent the median, and boxes represent the interquartile range (IQR) from the median