Table 5. Relative significance of features selected (descending order) for second-level regression models.

Regression Models	Feature removed	Percentage change
		PCC	MAE	RMSE
High	T × MCPhe	−3.96	39.83	36.49
	TP	−2.32	24.85	14.90
	VT	−1.98	22.43	26.52
Medium	WQ	−52.35	55.83	40.49
	ANC × MCAliphatic	−47.28	35.51	27.50
	ER	−35.52	26.77	16.13
	R × AbsCharge	−15.74	6.28	7.41
	VT	−2.04	13.28	3.71
	MCCys × pI	−4.90	6.88	3.49
Low	Y × nlogPFR	−36.06	15.54	17.17
	S × MCNPH	−2.54	11.31	5.08
	y × transmembrane	6.02	5.60	0.76
	F × logPFR	2.90	0.36	−2.14

Percentage changes were evaluated using LOOCV, with reference to the high, medium and low regression models trained using respective feature subsets.