glogP simple atom types + xlogP training set
Derive simple atom types by considering atomic number, aromaticity, charge,
valence, hydrogen count and substitution count. Yields 36 types from xlogP training
set. 1 type (=NH) occurs only once, leaving 35 types.
simple
| smarts |
train_freq |
coefficient |
error |
| [c+0v4H0D3] |
1471 |
0.294 |
0.019 |
| [c+0v4H1D2] |
1453 |
0.318 |
0.012 |
| [O+0v2H0D1] |
1075 |
-0.309 |
0.051 |
| [C+0v4H3D1] |
969 |
0.553 |
0.027 |
| [C+0v4H2D2] |
808 |
0.325 |
0.012 |
| [C+0v4H0D3] |
796 |
0.177 |
0.045 |
| [O+0v2H1D1] |
594 |
-0.401 |
0.034 |
| [O+0v2H0D2] |
543 |
-0.313 |
0.029 |
| [N+0v3H1D2] |
377 |
-0.523 |
0.040 |
| [N+0v3H2D1] |
371 |
-0.688 |
0.043 |
| [n+0v3H0D2] |
347 |
-0.375 |
0.035 |
| [C+0v4H1D3] |
283 |
-0.058 |
0.024 |
| [Cl+0v1H0D1] |
238 |
0.724 |
0.032 |
| [C+0v4H0D4] |
174 |
0.009 |
0.065 |
| [C+0v4H1D2] |
159 |
0.267 |
0.037 |
| [N+0v3H0D3] |
156 |
-0.893 |
0.052 |
| [F+0v1H0D1] |
140 |
0.405 |
0.033 |
| [N+0v5H0D3] |
138 |
0.631 |
0.106 |
| [N+0v3H0D2] |
129 |
0.317 |
0.062 |
| [C+0v4H0D2] |
108 |
0.573 |
0.160 |
| [n+0v3H1D2] |
103 |
-0.562 |
0.069 |
| [Br+0v1H0D1] |
87 |
0.964 |
0.058 |
| [n+0v3H0D3] |
87 |
-0.468 |
0.086 |
| [S+0v6H0D4] |
81 |
-0.216 |
0.118 |
| [N+0v3H0D1] |
79 |
-0.241 |
0.176 |
| [S+0v2H0D2] |
65 |
0.423 |
0.079 |
| [S+0v2H0D1] |
46 |
0.555 |
0.138 |
| [C+0v4H2D1] |
41 |
0.518 |
0.088 |
| [I+0v1H0D1] |
40 |
1.376 |
0.107 |
| [P+0v5H0D4] |
29 |
-0.225 |
0.156 |
| [s+0v2H0D2] |
13 |
0.663 |
0.188 |
| [o+0v2H0D2] |
12 |
-0.176 |
0.196 |
| [S+0v2H1D1] |
5 |
0.826 |
0.300 |
| [S+0v4H0D3] |
5 |
-1.174 |
0.305 |
| [C+0v4H1D1] |
4 |
0.020 |
0.372 |
| [N+0v3H1D1] |
1 |
|
|
|
|
Coefficient computation
- xlogp test set of 1850 compounds
- gNova CHORD count_match(smiles, smarts) function
- R statistical program linear models function
to correlate counts with experimental logP

Residual standard error: 0.6645 on 1817 degrees of freedom
Multiple R-Squared: 0.8104, Adjusted R-squared: 0.8068 F-statistic: 221.9 on 35 and 1817 DF, p-value: < 2.2e-16
|