TomkUCL
commented
7 months ago The process so far:
-
We are interested in retaining these cores whilst moving away from the N-oxide parent-compounds since the 51 purchased Enamine N-oxide set afforded no hits by ATPase/SPR assay at UNC. Therefore, we are diversifying the functional groups on either end of the diamine cores for docking.
-
Following our recent meeting concerning virtual libraries for the UCL cores, Peter Brown (@toluene44) went back and picked 316 diverse acids available from Enamine Building Blocks and created 10 libraries of about 93,000 compounds each. The link to download the folder is below. https://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.dropbox.com%2Fscl%2Ffo%2Flrf09mdh3qoe5k2006okp%2Fh%3Frlkey%3Dvwp7vounwat5h2l5zqmczdjf9%26dl%3D0&data=05%7C01%7Cthomas.knight.21%40ucl.ac.uk%7C389c7b54639143f4427008dbe07af024%7C1faf88fea9984c5b93c9210a11d9a5c2%7C0%7C0%7C638350590860482586%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=EUovSK4kxi86vPz5jjourKjKNnZA8I94KN8Z8LWdw0E%3D&reserved=0
-
The 10 rigid (non-linear alkyl) cores arose from the top 100 GLIDE scoring compounds from the de novo generated set, shown here:
-
I've filtered these compounds in DataWarrior using Lipinski filters and by removing 'nasty or toxic' functional groups to favour promising drug-like molecules as starting points. This process reduced the number of molecules from >93'000 per core to ~20'000 per core.
-
Next, I chose 1000 compounds based on 'structural diversity' such that compound 2 is more diverse than 1, and 3 is more diverse than 1 & 2, etc.
-
These 1000 'most diverse' compounds were then subjected to a virtual screen using the open-source virtual screening tool PyRx 0.8 with the AutoDock Vina 1.2 scoring function. References: Autodock Vina (https://vina.scripps.edu/, https://doi.org/10.1021/acs.jcim.1c00203, https://doi.org/10.1002/jcc.21334, https://doi.org/10.1021/acs.jcim.8b00545). PyRx 0.8 (https://pyrx.sourceforge.io/ , Small-Molecule Library Screening by Docking with PyRx. Dallakyan S, Olson AJ. Methods Mol Biol. 2015;1263:243-50., https://www.nature.com/articles/s41598-021-83626-x , https://www.nature.com/articles/s41598-020-60221-0 , https://doi.org/10.1016/B978-0-12-822312-3.00019-9)
- For the docking, the search grid was restricted to the RecA and 1B domains (containing ATP and RNA binding sites) to increase docking speed whilst exhaustiveness was set at 8 to maximise conformation exploration.
- I previously docked to apo-Nsp13 protomer (pdb 5rm9, i.e. not part of the replicase complex and in the absence of ATP or ssRNA).
- I wanted to try docking to the nsp13 engaged-RTC state (ATP/ssRNA bound pdb 7kro), since this is the closed form of the protein and molecules will be ATP/ssRNA competitive in vivo; this also fits the "molecular staple" idea that we discussed in the last meeting.
-
The helicase protein was prepared using Dock Prep in UCSF Chimera 1.6 and the ATP and ssRNA were removed from the structure to make these sites available for binding.
-
Exhaustiveness was run at 8 conformers per compound, then the compounds were ranked by their average binding energy. The poses were then visually inspected and validated using Biovia Discovery Studio 2021.
- After visual inspection of the top-scoring poses, the most promising poses will be sent to Geoff Wells to obtain~500 nanosecond AMBER molecular dynamics simulations.
mattodd
toluene44
Here I have described the current strategy regarding the retained diamine cores from the de-novo generated compounds since deprioritising the N-oxide series from which these cores originated.
Peter Brown enumerated these diamine cores using purchasable carboxylic acids from Enamine, giving ~93'000 compounds per core; @kipUNC has received this for virtual screening using Maestro GLIDE.
Simultaneously, I have been filtering these compounds based on their predicted physiochemical properties using Data Warrior https://doi.org/10.1021/ci500588j and then running a smaller-scale virtual screen (~1000 compounds/day) using open-source software PyRx 0.8 https://sourceforge.net/projects/pyrx/files/0.8/ , which uses the Autodock Vina scoring function to calculate free-binding energies, to prioritise for purchasing / synthesis. This process is described in more depth below.
A key question for prioritising compounds based on in-silico methods for the helicase which conformation of nsp13 should be used for docking?
Based on the 2022 paper from D Shaw's team "Ensemble cryo-EM reveals conformational states of the nsp13 helicase in the SARS-CoV-2 helicase replication–transcription complex" https://www.nature.com/articles/s41594-022-00734-6 . The four conformational states of the nsp13–RTC are:
These states are thought to regulate the RNA synthesis and proofreading of the SARS-CoV-2 virus.