Table 3.

Bootstrapping of the best two-feature combinations for the Linear, SVR and KNN regression models. In bootstrapping, the 20 data from the original dataset were randomly sampled with replacement. The regression models were generated 100 times and average value of the regression coefficients (r), RMSE and their standard deviations were calculated

	Two-features		r	RMSE
Linear	SCM_neg_H2	SASA_phobic_H3	0.56 ± 0.12	1.72 ± 0.42
SVR	SCM_pos_H2	SASA_phobic_Fv	0.87 ± 0.07	1.52 ± 0.29
KNN	SCM_pos_H2	SASA_phobic_Fv	0.90 ± 0.07	0.89 ± 0.22