Installation

The best way to install phold is using conda, as this will install Foldseek (the only non-Python dependency) along with the Python dependencies.

We would highly recommend installing conda via miniforge.

To install phold:

conda create -n pholdENV -c conda-forge -c bioconda phold

To utilise phold with GPU, a GPU compatible version of pytorch must be installed. By default conda will install a CPU-only version.

Therefore, if you have an NVIDIA GPU, please try:

conda create -n pholdENV -c conda-forge -c bioconda phold pytorch=*=cuda*

Pip

You can also install phold using pip.

pip install phold

You will need to have Foldseek (ideally v10.941cd33) installed and available in the $PATH.

Source

You can install the latest version of phold with potentially untested and unreleased changes into a conda environment using conda as follows:

conda create -n pholdENV pip foldseek python=3.13
conda activate pholdENV
git clone https://github.com/gbouras13/phold.git
cd phold 
pip install -e .

Mac (M1/M2/M3)

If you have a Mac that runs Apple Silicon (M1/M2/M3), please try:

mamba create -n pholdENV python==3.10
conda activate pholdENV
mamba install pytorch::pytorch torchvision torchaudio -c pytorch 
mamba install -c conda-forge -c bioconda phold

Intel XPU

If you have an Intel machine with an XPU, we have written separate installation instructions that show you how to install Phold.

pytorch

To utilise phold with GPU, a GPU compatible version of pytorch must be installed.

If it is not automatically installed via the installation methods above, please see this link for more instructions on how to install pytorch.

If you have an older version of the CUDA driver installed on your NVIDIA GPU, then you might find this link useful.

Phold has been tested on NVIDIA (A100, RTX4090), AMD (MI250) and Mac (M1 Pro) GPUs.

Installation on AMD GPUs requires a version of torch compatible with rocm e.g.

pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/rocm5.7

Database Installation

To download and install the phold database. Database downloads are multi-threaded (as of v1.0.0).

phold install -t <threads>

If you would like to specify a particular location for the database (e.g. if you use phold on a shared server), please use -d

phold install -d <path/to/databse_dir> -t <threads>
  • Note: You will need at least 8GB of free space (the phold databases including ProstT5 are 7.7GB uncompressed).

If you have an NVIDIA GPU available, you may wish to accelerate Foldseek using GPU. To do this, you will need to format the databases appropriately as follows

phold install -d <path/to/databse_dir> --foldseek_gpu -t <threads>

If you would like to install the Phold DB 3.16M (i.e. including 1.8M extra enVhog and efam proteins without PHROG assignment or functional annotation label), you can use --extended_db

Usage: phold install [OPTIONS]

  Installs ProstT5 model and phold database

Options:
  -h, --help             Show this message and exit.
  -V, --version          Show the version and exit.
  -d, --database TEXT    Specific path to install the phold database
  --foldseek_gpu         Use this to enable compatibility with Foldseek-GPU
                         acceleration
  --extended_db          Download the extended Phold DB 3.16M including 1.8M
                         efam and enVhog proteins without functional labels
                         instead of the default Phold Search 1.36M. Using the
                         extended database will likely marginally reduce
                         functional annotation sensitivity and increase
                         runtime, but may find more hits overall i.e.
                         including to efam and enVhog proteins that have no
                         functional labels.
  -t, --threads INTEGER  Number of threads  [default: 1]

Beginner Conda Installation

If you are new to using the command-line, please install conda using the following instructions.

First install some flavour of Anaconda.

There are lots of options but the best in our opinion is miniforge as this will automatically use the mamba solver, which has now been integrated into conda:

Please follow the instructions at the links to install based on your computer architecture.

After your installation is complete, you should add the following channels to your conda configuration:

conda config --add channels defaults
conda config --add channels bioconda
conda config --add channels conda-forge

We would recommend installing phold into a fresh environment. Assuming you installed miniforge, to create an environment called pholdENV with phold installed (assuming you have an NVIDIA GPU):

conda create -n pholdENV -c conda-forge -c bioconda phold pytorch=*=cuda*

If you don't have a GPU:

conda create -n pholdENV -c conda-forge -c bioconda phold

Then activate the environment

conda activate pholdENV

You can then check installation worked and download the phold databases:

phold -h
phold install

See the tutorial for more information on how to run phold.