pocketxmol¶

Summary

Source: https://github.com/pengxingang/PocketXMol
License: MIT
Path: /software/PocketXMol/
Documentation: https://github.com/pengxingang/PocketXMol

PocketXMol is an AI generative model that learns fundamental atom interactions, enabling applications governed by atom interactions within a pocket, including, small-molecule docking, peptide docking, and molecular conformation generation, structure-based drug design (SBDD), fragment linking/growing, PROTAC design, de novo linear/cyclic peptide design, and peptide inverse folding.

Using pocketxmol¶

pocketxmol is installed as a mamba environment under /software/PocketXMol/. For the setup have a look at /software/PocketXMol/sbatch.sh:

#!/bin/bash
#SBATCH --time=0-08:00:00
#SBATCH --constraint=A100|H100|H200
#SBATCH --partition=maxgpu
unset LD_PRELOAD
export PATH=/software/PocketXMol/pxm/bin:$PATH
export PXM=/software/PocketXMol
export PYTHONPATH=/software/PocketXMol/
export LD_LIBRARY_PATH=/software/PocketXMol/pxm/lib

mkdir -p configs/sample
cp /software/PocketXMol/configs/sample/*.yml configs/sample/
cp /software/PocketXMol/configs/sample/examples/dock_* configs/sample/
ln -s /software/PocketXMol/data data

python $PXM/scripts/sample_use.py \
        --config_task configs/sample/dock_smallmol.yml \
        --outdir outputs_examples \
        --device cuda:0

model-weights have been extracted into /software/PocketXMol/data. All data have been downloaded from zenodo to /software/PocketXMol/src/.

Installation¶

the installation used the following procedure:

cd /software
git clone https://github.com/pengxingang/PocketXMol.git

module load maxwell mamba
. mamba-init
mamba env create --prefix=/software/PocketXMol/pxm --file $PWD/environment_cu128_base.yml
mamba activate /software/PocketXMol/pxm
pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu128
pip install pyg_lib torch_scatter torch_sparse -f https://data.pyg.org/whl/torch-2.11.0+cu128.html
mamba install -c conda-forge libstdcxx-ng
pip install torch_geometric
pip install lightning
pip install torch-cluster