Table 2.

Comparison of predictive performance for AUC from three different approaches: RF, FRFL and FRF with two different model constructions using CCLE data.

Drug	Correlation			MAE
	RF	FRFL	FRF	RF	FRFL	FRF
Model parameters: #Tree = 150, m = 10, minimum leaf size = 10
Erlotinib	0.4408	0.4498	0.4641	0.0546	0.0541	0.0464
Nilotinib	0.3886	0.4318	0.4564	0.0465	0.0464	0.0376
PD-0325901	0.4716	0.5057	0.5658	0.1353	0.1335	0.1377
PLX-4720	0.2957	0.3137	0.4365	0.0494	0.0487	0.0396
TAE-684	0.2757	0.3385	0.3743	0.0728	0.0717	0.0684
Model parameters: #Tree = 500, m = 50, minimum leaf size = 5
Erlotinib	0.4381	0.4420	0.4701	0.0563	0.0557	0.0474
Nilotinib	0.4216	0.4393	0.4288	0.0470	0.0471	0.0391
PD-0325901	0.5928	0.5929	0.6381	0.1287	0.1282	0.1322
PLX-4720	0.3738	0.4195	0.5352	0.0492	0.0480	0.0393
TAE-684	0.3645	0.3888	0.4211	0.0711	0.0708	0.0679

For FRFL and FRF, node cost is calculated using 8 dose regions. Bold values indicate the best performances.