Installation

Our tools should be installed on a Linux system with Python3.8+.

Installing with a ‘package manager’

System Requirements

• Hardware Configuration:

  • NVIDIA GPU (VRAM ≥48GB, 16 - core CPU recommended)

• Software Environment:

  • CUDA 12.4+ driver

  • Python 3.8+ interpreter

  • Conda package management tool

We strongly recommend your installation executed in an isolated conda environment, so firstly run:

Quick Installation Process

1. Install the Rosetta Energy Calculation Suite

cd rosetta

# Download the Rosetta suite
wget https://downloads.rosettacommons.org/downloads/academic/3.14/rosetta_bin_linux_3.14_bundle.tar.bz2

# Extract the package
tar -xvf rosetta_bin_linux_3.14_bundle.tar.bz2

# Go back to the main directory
cd ..

2. Configure the Python Virtual Environment

# Create an isolated environment and activate it
conda create -n ACCESS python=3.8
conda activate ACCESS

# Install the PyTorch core framework
conda install pytorch torchvision torchaudio pytorch-cuda=12.4 -c pytorch -c nvidia

3. Install Core Dependencies

pip install argparse biopython torchdrug torch-scatter fair-esm scipy natsort

4. Compile the Geometric Vector Perceptor (GVP) Library

# Clone the official repository and apply the patch
git clone https://github.com/drorlab/gvp-pytorch.git

# Apply the patch
cd gvp-pytorch
git apply gvp_topology.patch
pip install .
cd ..