Computational drug repurposing screen

At the suggestion of @frankvondelft, I used the ensemble hybrid docking scheme we are using to dock Moonshot compounds for Folding@home free energy calculations (based on OpenEye OEDockingTK; code here) to dock the Broad repurposing library (updated 3/24/2020, downloadable here) to all fragments in the active site or dimer interface site.

The input and docked files are available here:

Top hits are listed in order (from best to worst) here:

Docked poses are here. That file contains a <fragments> SD tag with the name of the fragment structure to which the molecule was found to best bind to (e.g. x0104 ).

Interestingly, the top hit by a large margin is vonoprazan, a proton pump inhibitor used to treat GERD:

I’ve also docked Drugbank: CSV SDF

Edit: Linked to correct CSV file

Edit2: Updated with latest Broad screening compounds, and added Drugbank compounds.

Absolute alchemical free energy calculations on these poses are being spun up on Folding@Home today by Matthew Hurley in the Voelz lab!

Hi, thank you for sharing these results! I notice that some of the re-purposed drugs are charged (protonated or deprotonated) in your result set, but are not charged in the Broad data set. How did you determine which molecules to modify in this way, and what was the basis for the modification? Also, was the protease modified in this way as well?

(P.S. I think you linked to the incorrect .csv file in your post. I believe the ordered results are in a file named broad-repurposing-docked.csv, if I’m not mistaken.)

Thanks for the feedback, @LAntunes! I’ve corrected the link to the CSV file above.

The simple ensemble docking approach used the OpenEye toolkit method OEGetReasonableTautomers to enumerate all reasonable tautomers (and associated protonation states) but made no effort to penalize various tautomers or protonation states by their solution state penalties (e.g. using Epik). The protein was prepped with the OpenEye SpruceTK, but protonation and tautomeric states of the protein were not exhaustively enumerated. Much more detailed work can be done there using, for example, constant pH simulation methodologies or Monte Carlo enumeration schemes like MCCE2.