. 2017 Sep 12;7:11347. doi: 10.1038/s41598-017-11665-4

Table 6.

Mean Square Error (MSE) and Mean Absolute Error (MAE) between actual and predicted responses using 3 fold cross validation for different cancer types for multivariate case in CCLE dataset.

Cancer Types	Drug Names	Mean AUC of Cancer Types	MRF		mHARF
Cancer Types	Drug Names	Mean AUC of Cancer Types	MSE	MAE	MSE	MAE
CNS & Skin	AZD6244	0.09 & 0.30	0.0158	0.0982	0.0142	0.0878
	PD-0325901	0.13 & 0.43	0.0278	0.1309	0.0244	0.1184
	PLX4720	0.05 & 0.17	0.0105	0.0797	0.0096	0.0748
Skin & Ovary	17-AAG	0.46 & 0.36	0.0173	0.1037	0.0166	0.1012
	AZD0530	0.07 & 0.13	0.0046	0.0550	0.0048	0.0572
	AZD6244	0.30 & 0.11	0.0152	0.1028	0.0140	0.0884

Whenever, mean differences of the AUC distributions between two cancer types are higher, mHARF is doing significantly better than MRF, but in cases where mean difference of the AUC distributions between two cancer types are close, MRF and mHARF are performing similar.