Table 2. Summary of kNN models.

Models	Data splitting (training/testing)	Neighborsa	b	c	Descriptors
1	27/21	2	0.81	0.78	KierA2, GCUT_SLogP_0, radius
2	28/20	2	0.82	0.74	KierA2, BCUT_SLogP_3, diameter
3	33/15	2	0.75	0.83	PEOE_VSA_HYD, vsa_acc, radius
4	35/13	2	0.77	0.86	PEOE_VSA_HYD, SMR_VSA6, PEOE_PC+, radius
5	36/12	2	0.74	0.89	KierA2, b_ar, radius
6	37/11	4	0.76	0.93	KierA2, GCUT_SLogP_0, radius

^aOptimal number of nearest neighbors.

^bLOO cross-validated correlation coefficient for the training set.

^cCorrelation coefficient for the test set.