this page briefly describes a local installation of Alphafold 2.3.2 (no docker or singularity involved).
Setup
alphafold installations reside in /software/alphafold/<version>.
- have a look at alphafold.sh and sbatch-alphafold.sh in the installation folder of a specific version
- for alphafold >= 2.3.2 there is also a module, module load maxwell alphafold/2.3.2
- alphafold < 2.3.2 use /beegfs/desy/group/it/ReferenceData/alphafold as databases
- alphafold >= 2.3.2 use /beegfs/desy/group/it/ReferenceData/alphafold2.3 as databases
Examples
see alphafold.sh, sbatch-alphafold.sh as examples.
- for alphafold>=2.3.2 see alphafold-module.sh for an example using the module
Installation
tmpdir=/scratch/$USER inst_dir=/software/alphafold/2.3.2L pushd $tmpdir wget https://github.com/conda-forge/miniforge/releases/latest/download/Miniforge3-Linux-x86_64.sh sh Miniforge3-Linux-x86_64.sh -b -p $inst_dir export PATH=$inst_dir/bin:$PATH # create a mamba-init, conda-init . mamba-init mamba create -n af2.3.2 python=3.9 mamba activate af2.3.2 # # alphafold itself # wget https://github.com/deepmind/alphafold/archive/refs/tags/v2.3.2.tar.gz -O alphafold-2.3.2.tar.gz tar xf alphafold-2.3.2.tar.gz mv alphafold-2.3.2 $inst_dir/alphafold rm alphafold-2.3.2.tar.gz # deps mamba install -y -c bioconda hmmer hhsuite==3.3.0 kalign3 mamba install -y -c conda-forge openmm=7.5.1 pdbfixer pip # 7.5.1 required! mamba install -y -c nvidia cudatoolkit=11.8 cudnn==8.8 mamba install -y absl-py=1.0.0 biopython=1.79 chex=0.0.7 dm-haiku=0.0.10 dm-tree=0.1.8 immutabledict=2.0.0 pandas=2.0.3 python3 -m pip install --upgrade "jax[cuda11_pip]" -f https://storage.googleapis.com/jax-releases/jax_cuda_releases.html python3 -m pip install tensorflow-cpu==2.13.0 ml-collections==0.1.0 rm -rf /software/alphafold/2.3.2L/envs/af2.3.2/lib/python3.9/site-packages/jax* python3 -m pip install --upgrade "jax[cuda11_pip]" -f https://storage.googleapis.com/jax-releases/jax_cuda_releases.html popd # # patch # pushd $inst_dir/envs/af2.3.2/lib/python3.9/site-packages patch -p0 < $inst_dir/alphafold/docker/openmm.patch popd # # get stereo_chemical_props.txt # wget -P $inst_dir/alphafold/common/ https://git.scicore.unibas.ch/schwede/openstructure/-/raw/7102c63615b64735c4941278d92b554ec94415f8/modules/mol/alg/src/stereo_chemical_props.txt --no-check-certificate conda deactivate # setup alphafold.sh and T1050-2.fasta