| 1 |
Interpretation of machine learning models using shapley values: application to compound potency and multi‑target activity predictions |
2020 |
Journal of Computer-Aided Molecular Design |
predict compound potency & multi-protein active or not |
|
ECFP4 |
RF, GB, DNN(MT) |
SHAP |
|
|
| 2 |
DEEPScreen: high performance drug–target interaction prediction with convolutional neural networks using 2-D structural compound representations |
2020 |
Chem. Sci. |
predict 704 single-protein active or not |
Collected from ChEMBL |
compound 2D structural |
CNN & LR, RF, SVM(comparison) |
x |
in vitro experiment on JAK |
DEEPscreen. |
| 3 |
Machine Learning Models Based on Molecular Fingerprints and an Extreme Gradient Boosting Method Lead to the Discovery of JAK2 Inhibitors |
2019 |
Chem. Sci. |
predict IC50 on JAK2 |
PubChem, BindingDB, DUD-E and literature |
MACCS, ECFP and mol2vec |
XGBoost |
x |
in vitro experiment on JAK |
|
| 4 |
Profiling Prediction of Kinase Inhibitors: Toward the Virtual Assay |
2016 |
J. Med. Chem. |
predict ~200 kinase inhibitors |
in-house profiling, Tang, ChEMBL |
ECFP-like, FCFP-like |
Random forest |
x |
|
|
| 5 |
Out-of-the-box deep learning prediction of pharmaceutical properties by broadly learned knowledge-based molecular representations |
2020 |
Nature machine intelligence |
predict pharmaceutical properties |
PubChem & 26 pharmaceutically relevant benchmark datasets |
2D compound molecular descriptors |
CNN-based MolMapNet(GCN, GAT, CNN?) |
x |
|
MolMap |