. 2017 Jun 7;9:35. doi: 10.1186/s13321-017-0211-5

Table 7.

Statistical parameters for the best models for 2–6 variables for the physicochemical property log K, considering the 31 structures as the training set

Size	R ²	Q ²_loo	Q ²_boot	a (Q ²)	F	Models	Equations
2	0.778	0.734	0.738	−0.208	49.16	log K = 1.596 (±0.885) + 3.809 (±0.582)	(19)
						TS[1]_MX_B_AB_nCi_2_SS12_T_KA_a-h − 0.118 (±0.011)
						KH[1]_MX_F_AB_nCi_2_MP2_T_KA_h
3	0.863	0.826	0.820	−0.259	57.14	log K = −32.132 (±3.841) − 75.624 (±9.789)	(20)
						TS[1]_RA_F_AB_nCi_2_MP2_T_KA_h + 135.484 (±13.179
						TS[4]_PN_Q_AB_nCi_2_MP0_T_KA_h + 1782.101 (±257.835)
						KH[2]_PN_B_AB_nCi_2_SS8_T_KA_v-h
4	0.915	0.887	0.879	−0.324	70.59	log K = −66.472 (±6.939) − 0.223 ± 0.021)	(21)
						AC[2]_MX_B_AB_nCi_2_SS7_T_KA_r-h + 0.407 (±0.089)
						TS[5]_HM_B_AB_nCi_2_SS8_T_KA_v-h + 131.848 (±10.928)
						TS[4]_PN_Q_AB_nCi_2_MP0_T_KA_h + 3323.451 (±355.509)
						KH[2]_PN_B_AB_nCi_2_SS8_T_KA_v-h
5	0.932	0.902	0.890	−0.376	68.53	log K = −70.522 (±6.342) − 0.246 (±0.020)	(22)
						AC[2]_MX_B_AB_nCi_2_SS7_T_KA_r-h + 0.422 (±0.081)
						TS[5]_HM_B_AB_nCi_2_SS8_T_KA_v-h + 144.507 (±9.991)
						TS[4]_PN_Q_AB_nCi_2_MP0_T_KA_h + 4616.536 (±15.439)
						GV[2]_MX_Q_AB_nCi_2_MP3_X_KA_h + 3536.215 (±324.863)
						KH[2]_PN_B_AB_nCi_2_SS8_T_KA_v-h
6	0.942 (0.960)^a	0.914 (0.937)^a	0.898 (0.925)^a	−0.414 (−0.465)^a	65.26 (91.74)^a	log K = −81.005 (±6.216) − 0.233 (±0.020)	(23)
						AC[2]_MX_B_AB_nCi_2_SS7_T_KA_r-h − 39,144.250 (±4.757)
						AC[2]_MN_B_AB_nCi_2_MP2_A_KA_c-h + 0.572 (±17.485)
						TS[5]_HM_B_AB_nCi_2_SS8_T_KA_v-h + 120.683 (±1.681)
						TS[4]_PN_Q_AB_nCi_2_MP0_T_KA_h + 0.804 (±0.354)
						TS[6]_HM_Q_AB_nCi_2_SS0_A_KA_h + 3979.089 (±310.376)
						KH[2]_PN_B_AB_nCi_2_SS8_T_KA_v-h

^aCompound 31 excluded, taken as outlier, is not taken into account in the training set