Application of Multivariate Adaptive Regression Splines (MARSplines) for Predicting Hansen Solubility Parameters Based on 1D and 2D Molecular Descriptors Computed from SMILES String
Table 3
MARSplines HB(25, 2) model regression factors along with their weights.
Factor
ai ± SD
Mathematical relationships
F0
12.6280 ± 0.4535
F1
5.5560 ± 0.7079
max(0; SHBd-0.84757)
F2
−10.4070 ± 1.0496
max(0; 0.84757-SHBd)
F3
1.0900 ± 0.1333
max(0; 2.3406-CrippenLogP)
F4
0.0000 ± 0.0000
max(0; ECCEN-20.00000)·F2
F5
0.0000 ± 0.0000
max(0; 20.00000-ECCEN)·F2
F6
−4.0810 ± 0.4901
max(0; GATS2e-0.92565)
F7
−4.9500 ± 0.5455
max(0; 0.92565-GATS2e)
F8
−0.1460 ± 0.0470
max(0; WTPT-4-2.32775)
F9
−1.5640 ± 0.1466
max(0; 2.32775-WTPT-4)
F10
−62.8000 ± 7.3785
F1·max(0; SIC1-0.59306)
F11
−20.6450 ± 5.3855
F1·max(0; 0.59306-SIC1)
F12
21.0280 ± 3.0488
max(0; ETA_dEpsilon_D-0.05394)
F13
79.3130 ± 14.5139
max(0; 0.05394-ETA_dEpsilon_D)
F14
−0.3920 ± 0.0593
max(0; VE3_DzZ + 3.00162)·F8
F15
−88.4270 ± 13.1857
max(0; AATSC1i + 0.83463)·F13
F16
−100.3560 ± 19.2748
max(0; −0.83463-AATSC1i)·F13
F17
3.4670 ± 0.5511
max(0; AATSC7i-0.42042)
F18
3.1050 ± 0.6674
max(0; 0.42042-AATSC7i)
F19
0.1370 ± 0.0591
max(0; ATSC1v + 23.64635)·F12
F20
0.2160 ± 0.0470
max(0; −23.64635-ATSC1v)·F12
F21
1.8170 ± 0.7239
F2·max(0; AATSC2i + 0.09514)
F22
6.9340 ± 1.5981
F2·max(0; −0.09514-AATSC2i)
Model statistics: fitting criteria: N = 130, R2 = 0.974, = 0.970, F = 216.6, and LOF = 2.955; internal validation criteria: LMO (30%), Q2loo = 0.960, RMSE = 1.509, and MAE = 1.150.
