FAQ

Installation

The engine finds no GPU, or P2Rank does not start

See Troubleshooting on the installation page.

Can I use PocketVina without a GPU?

Docking needs an NVIDIA GPU. Reading results works on any computer, for example on a laptop with a copy of the results folder. This covers Results, show(), bust() and the commands info, query, show and export.

Runs

How do I resume a run that stopped?

Run the same command or code again. Proteins that are done are skipped, and all others are docked.

The run stops with "... holds another run"

A results folder holds one run with one set of ligands and settings. Use a new --out folder, or --force to replace the run.

The engine runs out of GPU memory

PocketVina retries a failed call with one engine less on that GPU. If the error CL_MEM_OBJECT_ALLOCATION_FAILURE or CL_OUT_OF_RESOURCES remains, lower --thread to no less than 1000, or use --procs-per-gpu 1.

Some ligands were not docked

This query lists the ligands that could not be prepared, and the reason:

pocketvina query --results results "SELECT ligand_id, status, error FROM ligands WHERE status <> 'ok'"

The engine docks only molecules made of C, N, O, S, P, F, Cl, Br and I.

Why do two runs give different poses?

With the same seed, ligands, settings and engine, a run gives the same poses again, on any number of GPUs. Changing --seed, --thread, the box or --ligands-per-call changes the poses. Different RDKit versions also build different 3D starting structures, so a pip and a conda environment can give different scores.

Which pocket was a ligand docked into?

The column best_pocket of best_scores() is the P2Rank rank of the pocket of the best pose. The pockets table has its centre, box size and residues (Results format).

Do I need to clear a cache after an update?

No. After a driver or PocketVina update, the engine compiles its GPU code into a new folder under ~/.cache/pocketvina/kernels/. You can delete the old folders. A folder set with POCKETVINA_KERNEL_CACHE is the exception: empty it yourself.

Viewing and analysis

The 3D page shows "Run this cell again to see the PocketVina 3D view."

Jupyter removes the page when a saved notebook is opened again. Run the cell again.

A pocketvina.viz view stays empty

These views load Mol* from the internet. Without internet access, use backend="py3dmol" or the 3D page of show() (3D views).

Every pose is invalid in PoseBusters

The column failed_checks names the checks that failed. Docked poses often fail on clashes with the protein or on the energy of the ligand. select(..., mode="all").bust() checks every mode of a pair, so you can find the modes that pass (PoseBusters checks).

Can I still get combined_results.csv?

Yes. This command writes it from a results folder:

pocketvina export --results results --what all --format csv --out tables