Table 6. Comparison of the Cross Validated Model Performance due to the Inclusion of Different Types of Features in the MDM and GNN Modelsa.
model | features | R2 | RMSE | Spearman |
---|---|---|---|---|
MDM | DFTb | 0.68 ± 0.02 | 1.23 ± 0.02 | 0.79 ± 0.01 |
Mol.b | 0.79±0.02 | 0.99±0.02 | 0.88±0.01 | |
Mol. + DFTb | 0.79±0.02 | 0.99±0.02 | 0.88±0.01 | |
MDM | 2Dc | 0.77±0.01 | 1.08±0.02 | 0.88±0.01 |
3Dc | 0.39 ± 0.01 | 1.76 ± 0.03 | 0.61 ± 0.01 | |
2D + 3Dc | 0.77±0.01 | 1.08±0.03 | 0.88±0.01 | |
2D + 3Datom-typec | 0.76 ± 0.01 | 1.10 ± 0.03 | 0.88 ± 0.01 | |
MDM | w/o MPb | 0.79 ± 0.01 | 1.04 ± 0.02 | 0.89 ± 0.00 |
with MPb | 0.79 ± 0.01 | 1.03 ± 0.02 | 0.89 ± 0.01 | |
MDM | w/o WSb | 0.79 ± 0.01 | 0.96 ± 0.04 | 0.89 ± 0.01 |
with WSb | 0.79 ± 0.01 | 0.96 ± 0.03 | 0.89 ± 0.00 | |
GNN | w/o 3D coordinatesb | 0.73±0.01 | 1.15±0.05 | 0.86±0.01 |
with 3D coordinatesb | 0.71 ± 0.01 | 1.20 ± 0.04 | 0.85 ± 0.00 | |
MetaLayer | 2Dc | 0.74±0.02 | 1.16 ± 0.04 | 0.87±0.00 |
3Dc | 0.71 ± 0.03 | 1.20 ± 0.07 | 0.85 ± 0.02 | |
2D + 3Dc | 0.74±0.02 | 1.15±0.04 | 0.87±0.00 |
2D denotes 743 2D descriptors and 59 molecular fragments. 3D denotes 37 3D descriptors. 3Datom-type denotes 3D descriptors containing atom type information. MP and WS represent melting point and Sterimol parameters.
Obtained using the entire data set.
Obtained using the molecules for which the relevant descriptors were available or able to be calculated.