Topic automatically created for discussing the designs at:
https://covid.postera.ai/covid/submissions/a1be81af-ddd9-4c01-8546-0400b9ee4a4e
Here are the PDB files containing the results of docking the submissions with 6YB7:
docked_6YB7_LUI-IND-a1b-1.pdb (205.8 KB) docked_6YB7_LUI-IND-a1b-2.pdb (205.8 KB) docked_6YB7_LUI-IND-a1b-3.pdb (206.0 KB)
Just read the rationale, interesting! I’m curious to see the results.
2 Likes
Thank you for your interest! I have created a similar submission that uses slightly more screening criteria, here: https://covid.postera.ai/covid/submissions/2c46affe-719d-43fe-ac03-722bb8f41f87
Also, if you have suggestions for changes that can be made to the structures that make them better candidates (e.g. easier to synthesize), I can evaluate them, and I’d be very grateful!
Just looked at it again; have you taken into account that there were covalent hits? LUI-IND-2c4-1 is suspicious, is the CH3C=O part of molecule next to the cysteine? Some designers didn’t realise that some compounds were actually covalently bound (protein-S-CH2-C(O)-molecule) and submitted methylketone compounds.
Smilarly, is one side of the piperazine of LUI-IND-2c4-3 expanding on the electrophilic part of the fragment hit? In which case it would clash with the protein.
Let me know if you understand, possible I am getting it wrong
Could you check and maybe submit these:
I altered slightly the atoms arrangement so they look chemically more makeable and have a chance to pass the filters
For: LUI-IND-2c4-1
CC(NC1=CN=C(CN2CCC(CC2)C3=CC=CC©=C3Cl)N=C1)=O
CC(NC1=CN=C(NC2CCC(CC2)C3=CC=CC©=C3Cl)N=C1)=O
CC(NC1=CN=C(CN2CCN(CC2)C3=CC=CC©=C3Cl)N=C1)=O
CC(NC1=CN=C(C(N2CCN(CC2)C3=CC=CC©=C3Cl)=O)N=C1)=O
CC(NC1=CN=C(C(N2CCC(CC2)C3=CC=CC©=C3Cl)=O)N=C1)=O
And for LUI-IND-2c4-2:
N#CC1=CC(C(C2=CC(O)=NC=C2)=O)=CC=C1N3[C@@H]4CCC[C@H]4OCC3
N#CC1=CC(S(C2=CC(O)=NC=C2)(=O)=O)=CC=C1N3[C@@H]4CCC[C@H]4OCC3
N#CC1=CC(C(C2=CC(O)=NC=C2)=O)=CC=C1N3CCOCC3
N#CC1=CC(S(C2=CC(O)=NC=C2)(=O)=O)=CC=C1N3CCOCC3
Hi, thanks so much for these suggestions! I have evaluated them by submitting them to docking with 6YB7 using AutoDock Vina (3 trials with an exhautiveness of 16). I have also added some with my own modifications, with the intention of making the compounds easier to synthesize. The results are displayed in the table below:
ID SMILES Binding Energy QED Score
1 CC(=O)NC1C=NC(NC2CCC(C3C=C(C)C=CC=3)CC2)=NC=1 -8.57 +/- 0.05 0.895 0.969
2 CC(=O)NC1C=NC(NC2CCC(C3C(Cl)=C(C)C=CC=3)CC2)=NC=1 -8.30 +/- 0.00 0.842 0.930
3 CC(=O)NC1C=NC(CN2CCC(C3C(Cl)=C(C)C=CC=3)CC2)=NC=1 -7.90 +/- 0.00 0.903 0.916
4 CC(=O)NC1C=NC(CN2CCC(C3C=C(C)C=CC=3)CC2)=NC=1 -7.67 +/- 0.12 0.939 0.908
5 CC(NC1=CN=C(NC2CCC(CC2)C3=CC=CC=C3Cl)N=C1)=O -7.70 +/- 0.42 0.871 0.890
6 CC(NC1=CN=C(CN2CCC(CC2)C3=CC=CC=C3Cl)N=C1)=O -7.50 +/- 0.00 0.922 0.889
7 CC(=O)NC1C=NC(CN2CCN(C3C=C(C)C=CC=3)CC2)=NC=1 -7.43 +/- 0.05 0.932 0.887
8 N#CC1=CC(S(C2=CC(O)=NC=C2)(=O)=O)=CC=C1N3[C@@H]4CCC[C@H]4OCC3 -7.63 +/- 0.05 0.864 0.882
9 CC(=O)NC1C=NC(CN2CCN(C3C(Cl)=C(C)C=CC=3)CC2)=NC=1 -7.37 +/- 0.09 0.909 0.875
10 N#CC1C(N2CCOCC2)=CC=C(S(=O)(=O)C2C=CN=C(C3CCCCC3)C=2)C=1 -7.87 +/- 0.05 0.763 0.872
11 CC(NC1=CN=C(CN2CCN(CC2)C3=CC=CC=C3Cl)N=C1)=O -7.23 +/- 0.05 0.922 0.867
12 N#CC1C=C(S(=O)(=O)C2OC3C=C(O)C=C(O)C=3C(=O)C=2)C=CC=1N1CCOCC1 -8.17 +/- 0.05 0.638 0.859
13 CC(NC1=CN=C(C(N2CCC(CC2)C3=CC=CC=C3Cl)=O)N=C1)=O -7.10 +/- 0.00 0.914 0.854
14 CC(NC1=CN=C(C(N2CCN(CC2)C3=CC=CC=C3Cl)=O)N=C1)=O -7.10 +/- 0.00 0.907 0.852
15 N#CC1C=C(S(=O)(=O)C2OC3C=CC=CC=3C(=O)C=2)C=CC=1N1CCOCC1 -7.97 +/- 0.05 0.670 0.852
16 N#CC1=CC(S(C2=CC(O)=NC=C2)(=O)=O)=CC=C1N3CCOCC3 -7.10 +/- 0.00 0.895 0.848
17 N#CC1=CC(C(C2=CC(O)=NC=C2)=O)=CC=C1N3[C@@H]4CCC[C@H]4OCC3 -7.17 +/- 0.09 0.857 0.843
18 N#CC1C=C(S(=O)(=O)C2OC=CC(=O)C=2)C=CC=1N1CCOCC1 -7.23 +/- 0.05 0.822 0.837
19 N#CC1C=C(S(=O)(=O)C2C=COC(=O)C=2)C=CC=1N1CCOCC1 -7.20 +/- 0.00 0.822 0.835
20 N#CC1=CC(C(C2=CC(O)=NC=C2)=O)=CC=C1N3CCOCC3 -6.83 +/- 0.05 0.869 0.819
21 N#CC1C(N2CCOCC2)=CC=C(S(=O)(=O)C2C=CN=CC=2)C=1 -6.70 +/- 0.00 0.851 0.802
The Score is a weighted sum of the Binding Energy (70%) and the QED (30%). In terms of derivatives of LUI-IND-2c4-1, the hydrazine functional group was not necessary, and neither was the ortho-Chlorine atom. But the meta-methyl group on the benzene ring did result in a large effect, in terms of the docking binding energy. Substituting the methylene “bridge” with a carbonyl resulted in lower binding affinity. The best candidate (#1), with a weighted score of 0.969, reflects these observations. In terms of LUI-IND-2c4-2, replacing the sulfone functional group with a ketone resulted in noticeably worse binding affinity. Simplifying the morpholine group resulted in a lower binding affinity. Replacing the 2-pyranone ring with a pyridine ring resulted in a slightly lower binding affinity. Replacing the 2-pyranone with a 4-pyranone improved the binding energy by a small amount, but it is unclear if this leads to an easier synthesis. Noteworthy is candidate #12, which was inspired by the Quercetin core. However, it is not considered very drug-like, according to the QED score.
Overall, I think candidate #1 is definitely worth submitting. I’d be curious how synthetically accessible candidate #12 is.
Just looked at it again; have you taken into account that there were covalent hits? LUI-IND-2c4-1 is suspicious, is the CH3C=O part of molecule next to the cysteine?
I haven’t looked in detail at the binding modes discovered by the docking algorithm, so it is possible that there are some unfavourable steric interactions. When I generated potential candidates, my algorithm treated the active site covalent hits as simply non-covalent. So it is possible that this is reflected in the submissions.
If you’d like, I can upload the .PDB files containing the docked molecules with 6YB7 for LUI-IND-2c4.