Submission Details
Molecule(s):
O=C(NCc1cccc(-c2cccc(=O)[nH]2)c1)C1=CC=C=CC1Oc1cccc(F)c1
Cc1cccnc1NC(=O)c1cccc(-c2ccc3c(c2)NNN=C3)c1
Nc1ccc(CN(Nc2cccc(N)c2)c2cccc(/C(F)=C/C=C3\N=CNCO3)c2)cc1
Design Rationale:
We used deep learning, or more specifically, variational autoencoders to generate new ligands for the SARS-CoV-2 Mpro. We used existing HIV and other drugs from ChEMBL as input for the autoencoder. The autoencoder then learned the pattern of the SMILES of the input and generated new ligands. The three new ligands we are submitting have better binding affinities than darunavir, tipranavir, colchicine, hydroxychloroquine, chloroquine, remdesivir, and favipiravir on the Mpro. PDB ID: 6LU7 Active site: x = -11.70, y = 13.90, z = 70.55 (r = 12) More information in the paper: https://chemrxiv.org/articles/On_the_Generation_of_Novel_Ligands_for_SARS-CoV-2_Protease_and_ACE2_Receptor_via_Constrained_Graph_Variational_Autoencoders/12011157
Other Notes:
We didn't use any of the fragments. Since our PDB files are compressed in a .zip file, here is the url of the file: https://drive.google.com/open?id=11vohoRCJ8--kTwo02fV9kL9toCtUKZJP