How accurately can one predict drug binding modes using AlphaFold models?
- Biophysics Program, Stanford University, United States
- Department of Computer Science, Stanford University, United States
- Department of Molecular and Cellular Physiology, Stanford University School of Medicine, United States
- Department of Structural Biology, Stanford University School of Medicine, United States
- Institute for Computational and Mathematical Engineering, Stanford University, United States
Figures
Figure 1 with 2 supplements
Structural accuracy of modeled binding pockets.
Binding pockets are defined to include all amino acid residues with any atom within 5 Å of the ligand in an experimentally determined structure. We compute all-atom binding pocket root mean squared …
Figure 1—figure supplement 1
Structural accuracy of modeled proteins.
We compute an all-atom root mean squared deviation (RMSD) between each modeled structure and all experimentally determined structures of the same protein. We also compute an all-atom RMSD between …
Figure 1—figure supplement 2
Structural accuracy of modeled binding pockets.
This figure is similar to Figure 1 but includes an additional bar for the case where each ligand is docked only to protein structures determined experimentally in complex with a ligand very …
Figure 2 with 2 supplements
Accuracy of ligand-binding poses predicted by computational docking to AlphaFold 2 models, traditional template-based models, or protein structures determined experimentally in complex with a ligand different from the one being docked.
We plot the fraction of docked ligands whose pose is predicted correctly (see Methods). Error bars are 90% confidence intervals calculated via bootstrapping. *** for p-values <0.001, ns for p-values …
Figure 2—figure supplement 1
Accuracy of ligand-binding poses predicted by computational docking to AlphaFold 2 models, traditional template-based models, or experimentally determined protein structures.
This figure is identical to Figure 2 apart from the addition of two bars: (1) The red bar (docked to reference structure) is for the case where one docks the ligand from an experimentally determined …
Figure 2—figure supplement 2
Accuracy of ligand-binding poses predicted by computational docking to AlphaFold 2 models, traditional template-based models, or protein structures determined experimentally in complex with ligands different from the one being docked or very different from the one being docked.
These plots are similar to those of Figure 2 and Figure 2—figure supplement 1, except that docking is performance with (A) Glide Dock XP and (B) Rosetta docking. Error bars are 90% confidence …
Figure 3 with 1 supplement
An example in which docking to an AlphaFold 2 (AF2) model yields poor results even though the model’s binding pocket has high structural accuracy.
We predict the binding pose of the drug aprepitant to its target, the neurokinin-1 receptor (NK1R) given either the AF2 model (orange) of NK1R or the experimentally determined structure (blue, PDB …
Figure 3—figure supplement 1
An example in which docking to an AlphaFold 2 (AF2) model yields poor results even though the model’s binding pocket has high structural accuracy.
We predict the binding pose of the drug-like ligand A-77636 to the D1 dopamine receptor (DRD1) given either the AF2 model (orange) of DRD1 or the experimentally determined structure (blue, PDB …
Figure 4
An example in which docking to a traditional template-based model yields better results than docking to an AlphaFold 2 (AF2) model, even though the AF2 model’s binding pocket has higher structural accuracy.
We predict the binding pose of the psychedelic LSD to its primary target, the serotonin 2A receptor (5HT2A) given either the AF2 model (orange) or a traditional model (green) of 5HT2A. (A, B) The …
Figure 5 with 1 supplement
Pose prediction accuracy as a function of binding pocket structural accuracy when docking to AlphaFold 2 (AF2) models or experimentally determined structures.
Docking to an experimentally determined structure generally leads to more accurate pose prediction than docking to an AF2 model with the same binding pocket root mean squared deviation (RMSD). The …
Figure 5—figure supplement 1
Pose prediction accuracy as a function of binding pocket structural accuracy when docking to AlphaFold 2 (AF2) models, experimentally determined structures or traditional template-based models, with different docking protocols.
Using our docking data, we estimate how the pose prediction accuracy of ligand docking varies with the binding pocket all-atom root mean squared deviation (RMSD) of the structure being docked into …
Tables
Appendix 1—key resources table
| Reagent type (species) or resource | Designation | Source or reference | Identifiers | Additional information |
|---|---|---|---|---|
| Software | Schrödinger tools | Schrödinger | Version: 2021-1 glide-v9.0 | Glide, Maestro, Protein Preparation Wizard |
| Software | PyMOL | Schrödinger | PyMOL 3.8 | |
| Software | AlphaFold 2.0.1 | https://doi.org/10.1038/s41586-021-03819-2 | ||
| Software | Rosetta GALigandDock | https://doi.org/10.1021/acs.jctc.0c01184 |
Additional files
-
Supplementary file 1
The full list of proteins and structures.
- https://cdn.elifesciences.org/articles/89386/elife-89386-supp1-v1.docx
-
MDAR checklist
- https://cdn.elifesciences.org/articles/89386/elife-89386-mdarchecklist1-v1.pdf
