My initial compound library had 348,276 compounds. After applying various filters like functional group filters, Physio-chemical filters, removing duplicate compounds, removing Pan Assays Interferences Compounds etc my final size of compound library has reduced by about 90 percent. I have applied these filters from FAF-Drugs server (http://mobyle.rpbs.univ-paris-diderot.fr) using Leadlike soft Physiochemical filters, functional group filters available there and all the three PAINS filters (A,B and C).
But 34000 compounds is still huge number for virtual screening (VS). My question is: are there any other online filters (freely available would be good!!) that i can apply to reduce the number?
If I proceed with these numbers for VS it would generate huge data. Therefore, I need your expert opinion on efficient post virtual screening analysis. What tools are available out there that can be used to analyses and interpret huge virtual screening data .Name of software or links of papers describing such work will also be a great start for me.
hello Rohan Ji ,
I have done HTVS for more than 1L compounds in Maestro Schrodinger s/w. 34000 compounds is not a big deal. Can I know what kind of s/w or program do u use for VS.? so that I can suggest u some ideas. Thanks
Dear kishna I have planned using autodock for VS. I expect that compound number will again drop-down during VS. My concern is regarding proper analysis and presentation.
my suggestion to u is tht u can use Schrodinger , so it will be easy for u to screen the compounds and have better info analysis.....
Hello Rohan,
I agree with Kishna in the sense that 34k is really shouldn't be an issue for Autodock!
I'm regularly doing VS with Autodock on the scale of ~70k, and it's not taking forever neither! :) Good luck!
You are right to be concerned about the post-docking analysis. This has to be done well to generate the best possible results. Relying on docking score only to choose your final set for screening will not work well. I would recommend thinking about the following filtering strategies
1) Removal of structures which have bad clashes with protein or poor docked geometry
2) promotion of structures that have good consensus scores when rescored with a second (and third./fourth) scoring function
3) promotion of structures which make the interactions with the active site that you think important
4) Clustering by chemistry (e.g. Murcko scaffolds) followed by chemical diversity analysis to generate a diverse subset.
Post docking analysis tools exist which are designed to do most of 1-3 (e.g. GoldMine which comes with GOLD). i've used the free pipelining tool Desktop Knime to do step 4.
- Badges
- Science topic
- Similar topics
- Chemistry
- Biochemistry