VEGF receptors (VEGFRs) are tyrosine kinase receptors that are activated by vascular endothelial growth factor (VEGF). VEGF is a signaling protein involved in both vasculogenesis (the formation of the circulatory system) and angiogenesis (the formation of new blood vessels). Aberrant angiogenesis is a symptom of a number of diseases, including cancer, and so inhibitors of VEGFR are of therapeutic interest.
Basic (✓): Pharmacophore search with Pharmit
Construct an effective pharmacophore for identifying potential VEGFR inhibitors. We will use as a starting point PDB 4ASE, which is VEGFR2 bound to tivozanib (ligand code AV9).
- Load 4ASE into Pharmit.
- IMPORTANT: change the library to screen to the DUD-E VEGFR benchmark, a contributed library.
- Refine the pharmacophore by removing, adding or modifying features so that screening this library identifies an enriched set of actives. The enrichment factor (EF) and F1 score are shown in the bottom right. If you'd like, you can use other VEGFR structures as starting points.
- Minimize your results.
- Once you have a top hit with a score less than -12 and an mRMSD under 2, take a screenshot of the full Pharmit session that shows the top hit, its score, and the pharmacophore.
Upload your screenshot to GradeScope.
Extra (✓+): Scoring and docking with gnina
Take your top hit from Pharmit (minimized_results.sdf.gz) and use gnina on the cluster to rescore and dock it. You also need the receptor, rec.pdb: the 4ASE protein only, with the ligand and waters removed.
Connect interactively to a GPU node on the cluster. If you get errors about CUDA libraries being missing, run module load cuda.
First, minimize the Pharmit pose in the receptor. This outputs the CNNscore (probably a good pose) and the CNNaffinity (pK estimate, where 9 is 1 nM) for the ligand in the provided pose:
/net/galaxy/home/koes/dkoes/local/bin/gnina -r rec.pdb -l minimized_results.sdf.gz --minimize -o min.sdf
Next, perform a full docking, using the minimized ligand to define the search box:
/net/galaxy/home/koes/dkoes/local/bin/gnina -r rec.pdb -l min.sdf --autobox_ligand min.sdf -o dock.sdf
Does docking find the same pose identified through the pharmacophore search? /net/galaxy/home/koes/dkoes/local/bin/obrms is a useful utility for answering this question quantitatively.
Upload to GradeScope a PyMOL figure showing both poses (the Pharmit pose and the top docked pose), the CNNscore and CNNaffinity of each, the RMSD between the poses, and your answer to the question above.