Table 1.

Fivefold Cross validation results for GDSC dataset response prediction for four drug sets (S₁, S₂ and S₃ sets consists of highly correlated responses whereas the set of S₄ has low correlated responses)

			Correlation coefficients
Drug set	Correlation	Drug name	EN	KBMTL	RF	MRF
	among responses
S1	0.8439	RDEA119	0.62	0.57	0.63	0.66
		PD-0325901	0.48	0.47	0.61	0.63
S2	0.8410	BI-2536	0.23	0.23	0.26	0.28
		GW843682X	0.30	0.28	0.31	0.33
S3	0.8366	CI-1040	0.46	0.51	0.59	0.60
		PD-0325901	0.50	0.52	0.62	0.65
S4	6.59e-7	Axitinib	0.31	0.33	0.36	0.32
		Mitomycin C	0.28	0.25	0.37	0.38

Correlation coefficients between actual and predicted drug responses using Elastic Net (EN), Kernelized Bayesian multitask learning (KBMTL), Random Forest (RF) and Multivariate Random Forest (MRF) are reported. The parameters for EN and KBMTL were same as the parameters used by earlier drug sensitivity prediction studies of Barretina and et al. (2012) and Gonen and Margolin (2014) respectively.