New COVID-19 protease inhibitors

designed by AI with recurrent neural networks, molecular docking, and QSAR

Information about generated molecule # 43037

O=C(N[C@H](CS(=O)(=O)c1ccc2ccccc2c1)C(=O)N[C@H]1c2ccccc2OC[C@H]1O)O[C@H]1CCOC1



Predicted properties

Property Value Comments
Docking score -9.1 VINA score, less is better
pIC50 (SARS, M) 5.28 bigger is better
pIC50 (HIV, M) 9.84 bigger is better
logS (M) -3.71 Solubility based on the Transformer Model, bigger is better.
Cycle >4 The cycle when the molecule was generate, the bigger cycle the more RNN is inventive.
Cardiotoxicity, hERG 0.0002 Probability of being hERG binder (0-1), less is better.
Mutagenicity, AMES 0.2486 Probability of being mutagenic (0-1), less is better.
Caco-2 permeability -5.8288 log(sm/s)
HIA, intestinal absorbtion 99.9 %
LD50, oral (rats) 2.5 -log(M)
SARS-CoV inhibition 22.0 %
Molecular weight 540.59 g/mole
This molecule was not found in references databases.

Generator summary

Random prospective compound




O=C(NCc1nnn[nH]1)OCC1c2ccccc2-c2ccccc21

COVID-19 protease

The structure has been recently solved, PDB code 6lu7. 6ul7 structure
Developed in group of Cheminformatics in the Institute of Structural Biology at Helmholtz-Zentrum Muenchen, Tetko group
With support of BIGCHEM GmBH (bigchem.de) and RULIS Ltd. (rulis.ru).