How can we select the best ligands among more than 60K compounds for a protein based on their binding affinities?
These compounds are from an antiviral library and I performed molecular docking using autodock vina.
Should I select by ranking them on the basis of high to low binding affinity and then taking the top 10 or 100 ligands from it?
There could be multiple criteria to select the "best" ligand after docking such as binding affinity (suggested by autodock, and you already noted this), binding conformation, binding site (active site, allosteric site, interface, or your desired site), bound to know amino acids that participate in protein function to name a few (from top of my head). I would also include the visual inspection of whether the lowest (or selected ligands) binding conformation is sensible.
You should search the literature for the possible active site of your target, and narrow down your criteria to include only those that satisfy this.
I hope you have already read this, but just for additional information, you can take guidance from this article as well.
https://www.nature.com/articles/s41596-021-00597-z
Good luck.
Ayaz Anwar From literature studies I was able to understand that the selection of final hit compounds, depends upon the kind of docking tool used and the protein structure (whether the protein has the literature-known binding site available or a pre-docked ligand/drug bind to it to act as a template for best-docked score).
In my case, the protein has no known binding sites and is computationally determined in various literature with different docking platforms. Mine is done via auto dock vina (since it is freely available) with a different set of compound libraries (from the ChemDiv database). After pinpointing the common residues found in various literature, I was able to make out the active site binding residues for my protein. Afterwards, I performed rigid docking with 60K compounds. From there onwards, I am facing problems with how to select the best hit compounds.
For residue-ligand interactions, I need to have a small pool of protein-ligand complexes to actually check if the ligand is binding to the correct residues (as my computational resource as well as m, as well as my timeline, is limited) for that I thought to select the best compounds based on binding affinity.
Therefore, is this the correct methodology to take a certain binding affinity value to act as a cut-off and then proceed with physiochemical properties to filter out best-hit compounds?
Thanks
Hi,
1) VINA is fairly good at docking.
2) You should also try DoGSiteScorer, PockDrug, and other relevant resources to predict the binding site residues. This can give you a better clue, plus it also predicts the druggness of the site.
Therefore, is this the correct methodology to take a certain binding affinity value to act as a cut-off and then proceed with physiochemical properties to filter out best-hit compounds? "YES (with caution); as the binding affinity is crude and not rigorous"
You can also perform clustering to see where the highest number of ligands bound with protein and it aligns with your predicted active/binding site.
- Badges
- Science topic
- Similar topics
- Biological Science
- Biochemistry
- Enzymology
- Enzymes