AlphaFold 3 Integration for OneScience

This is the AlphaFold 3 implementation integrated into the OneScience framework as a submodule.

Overview

AlphaFold 3 is a state-of-the-art machine learning model for predicting protein structure, developed by DeepMind. This submodule integrates AlphaFold 3 into the OneScience framework, allowing it to be used as part of larger scientific computing workflows.

Installation

As Part of OneScience

The recommended way to install alphafold3 is as part of the complete OneScience package:

# Install OneScience with alphafold3 support

# install jackhmmer
# mkdir ~/hmmer_build ~/hmmer
# wget http://eddylab.org/software/hmmer/hmmer-3.4.tar.gz --directory-prefix ~/hmmer_build
# cd ~/hmmer_build &&  tar zxf hmmer-3.4.tar.gz && rm hmmer-3.4.tar.gz
# patch -p0 < jackhmmer_seq_limit.patch
# cd ~/hmmer-3.4
# ./configure --prefix ~/hmmer
# make -j && make install && cd ./easel && make install
# rm -R ~/hmmer_build

# # install extension
# pip install .[bio] -c constraints.txt
# cp -r /public/onestore/onedatasets/alphafold3/_dep xxx/
# export ALPHAFOLD3_DEP_DIR=/public/onestore/onedatasets/alphafold3/_dep
# cd src/onescience/flax_models/alphafold3/
# python build_extension.py

# optional create mmseqs2 database (please contact ai4s@sugon.com for mmseqs2 program)
export mmfasta=/root/public_databases
cd /root/public_databases && mkdir mmseqsDB
export mmdb=/root/public_databases/mmseqsDB
export CUDA_VISIBLE_DEVICES=0
mmseqs createdb $mmfasta/bfd-first_non_consensus_sequences.fasta $mmdb/small_bfd_db --gpu 1 --threads 32 --createdb-mode 2
mmseqs createdb $mmfasta/mgy_clusters_2022_05.fa $mmdb/mgnify_db --gpu 1 --threads 32 --createdb-mode 2
mmseqs createdb $mmfasta/uniprot_all_2021_04.fa $mmdb/uniprot_cluster_annot_db --gpu 1 --threads 32 --createdb-mode 2
mmseqs createdb $mmfasta/uniref90_2022_05.fa $mmdb/uniref90_db --gpu 1 --threads 32 --createdb-mode 2

Usage

# Import alphafold3 as part of onescience
import flax_model.alphafold3 as af3

# Access alphafold3 components
from flax_model.alphafold3 import structure, model, data

# Use alphafold3 functionality
print(f"AlphaFold3 version: {af3.__version__}")

Requirements

Python 3.11+
JAX with CUDA support (optional, for GPU acceleration)
CMake 3.28+ (for building C++ extensions)
Additional dependencies listed in pyproject.toml

License

This code is licensed under CC BY-NC-SA 4.0. See the original AlphaFold 3 repository for more details on usage restrictions and licensing terms.

Citation

If you use this code in your research, please cite the AlphaFold 3 paper:

Abramson, J., Adler, J., Dunger, J. et al. Accurate structure prediction of biomolecular interactions with AlphaFold 3. Nature 630, 493�?00 (2024).

OneScience
/

AlphaFold3