Skip to content

AlphaFold

AlphaFold je sistem umetne inteligence (AI), ki ga je razvil DeepMind in predvideva 3D-strukturo proteina na podlagi njegovega aminokislinskega zaporedja. Redno dosega natančnost, ki je konkurenčna eksperimentu.

Licenca

To ni uradno podprt Googlov izdelek.

Avtorske pravice 2022 DeepMind Technologies Limited.

Licenca za kodo AlphaFold

Licenciran pod licenco Apache, različica 2.0 ("licenca"); te datoteke ne smete uporabljati, razen v skladu z licenco. Kopijo licence lahko dobite na: https://www.apache.org/licenses/LICENSE-2.0.

Razen če to zahteva veljavna zakonodaja ali če ni pisno dogovorjeno, se programska oprema, distribuirana v skladu z licenco, distribuira na podlagi "KOT TAKA", BREZ KAKRŠNIH KOLI JAMSTEV ALI POGOJEV, izrecnih ali implicitnih. Glejte licenco za dovoljenja in omejitve, ki urejajo določen jezik v okviru licence.

Namestitev

AlphaFold je na voljo v gruči kot modul ali kot vsebnik Singularity.

Baze podatkov so v /ceph/hpc/software/alphafold.

Vsebniki singularnosti so že na voljo v standardnem imeniku:

ls  /ceph/hpc/software/containers/singularity/images/alpha*

Izpis:

/ceph/hpc/software/containers/singularity/images/alphafold-2.2.4.sif
/ceph/hpc/software/containers/singularity/images/alphafold-2.3.2.sif

Primer z vsebnikom:

export SING=/ceph/hpc/software/containers/singularity/images/alphafold-2.2.4.sif
export ALPHAFOLD_DATA_DIR=/ceph/hpc/software/alphafold

singularity run --env TF_FORCE_UNIFIED_MEMORY=1,XLA_PYTHON_CLIENT_MEM_FRACTION=4.0,OPENMM_CPU_THREADS=14,ALPHAFOLD_DATA_DIR=/ceph/hpc/software/alphafold \
  -B /ceph/hpc/software/alphafold \
  --pwd /app/alphafold \
  --nv $SING\
  --uniprot_database_path=${ALPHAFOLD_DATA_DIR}/uniprot/uniprot.fasta \
  --pdb_seqres_database_path=${ALPHAFOLD_DATA_DIR}/pdb_seqres/pdb_seqres.txt \
  --uniref90_database_path=${ALPHAFOLD_DATA_DIR}/uniref90/uniref90.fasta \
  --mgnify_database_path=${ALPHAFOLD_DATA_DIR}/mgnify/mgy_clusters_2018_12.fa \
  --template_mmcif_dir=${ALPHAFOLD_DATA_DIR}/pdb_mmcif/data_dir \
  --obsolete_pdbs_path=${ALPHAFOLD_DATA_DIR}/pdb_mmcif/obsolete.dat \
  --bfd_database_path=${ALPHAFOLD_DATA_DIR}/bfd/bfd_metaclust_clu_complete_id30_c90_final_seq.sorted_opt_a3m.ffdata \
  --pdb70_database_path=${ALPHAFOLD_DATA_DIR}/pdb70 \
  --uniclust30_database_path=${ALPHAFOLD_DATA_DIR}/uniclust30/uniclust30_2018_08 \
  --use_gpu_relax=True \
  --fasta_paths=/ceph/hpc/home/user/af_test/actn.fasta --max_template_date=2020-05-14 --model_preset=monomer --data_dir=$ALPHAFOLD_DATA_DIR--db_preset=full_dbs --output_dir=$PWD"$@"

ali v SBATCH

#!/bin/bash
#SBATCH -p gpu
#SBATCH --gres=gpu:4
#SBATCH -N 1
#SBATCH -c 8
#SBATCH -t 10:00:00

export SING=/ceph/hpc/software/containers/singularity/images/alphafold-2.2.4.sif
export ALPHAFOLD_DATA_DIR=/ceph/hpc/software/alphafold

singularity run --env TF_FORCE_UNIFIED_MEMORY=1,XLA_PYTHON_CLIENT_MEM_FRACTION=4.0,OPENMM_CPU_THREADS=14,ALPHAFOLD_DATA_DIR=/ceph/hpc/software/alphafold \
  -B /ceph/hpc/software/alphafold \
  --pwd /app/alphafold \
  --nv $SING\
  --uniprot_database_path=${ALPHAFOLD_DATA_DIR}/uniprot/uniprot.fasta \
  --pdb_seqres_database_path=${ALPHAFOLD_DATA_DIR}/pdb_seqres/pdb_seqres.txt \
  --uniref90_database_path=${ALPHAFOLD_DATA_DIR}/uniref90/uniref90.fasta \
  --mgnify_database_path=${ALPHAFOLD_DATA_DIR}/mgnify/mgy_clusters_2018_12.fa \
  --template_mmcif_dir=${ALPHAFOLD_DATA_DIR}/pdb_mmcif/data_dir \
  --obsolete_pdbs_path=${ALPHAFOLD_DATA_DIR}/pdb_mmcif/obsolete.dat \
  --bfd_database_path=${ALPHAFOLD_DATA_DIR}/bfd/bfd_metaclust_clu_complete_id30_c90_final_seq.sorted_opt_a3m.ffdata \
  --pdb70_database_path=${ALPHAFOLD_DATA_DIR}/pdb70 \
  --uniclust30_database_path=${ALPHAFOLD_DATA_DIR}/uniclust30/uniclust30_2018_08 \
  --use_gpu_relax=True \
  --fasta_paths=/ceph/hpc/home/user/af_test/actn.fasta --max_template_date=2020-05-14 --model_preset=monomer --data_dir=$ALPHAFOLD_DATA_DIR--db_preset=full_dbs --output_dir=$PWD"$@"

Razpoložljivi moduli:

module spider Alpha

Izpis:

----- /cvmfs/sling.si/modules/el7/modules/all ----------------------
   AlphaFold/2.1.2-foss-2021a
   AlphaFold/2.2.2-foss-2021a-CUDA-11.3.1
   AlphaFold/2.3.0-foss-2021b-CUDA-11.4.1
   AlphaFold/2.3.4-foss-2022a-CUDA-11.7.0-ColabFold (D)

Primer z modulom za particijo GPU:

#!/bin/bash
#SBATCH -p gpu
#SBATCH --gres=gpu:4
#SBATCH -N 1
#SBATCH -c 8
#SBATCH -t 10:00:00

module load AlphaFold/2.2.2-foss-2021a-CUDA-11.3.1

alphafold \
--data_dir=/ceph/hpc/software/alphafold/ \
--fasta_paths=$(pwd)/sample.fasta \
--output_dir=$(pwd) \
--model_preset=multimer \
--db_preset=full_dbs \
--use_gpu_relax=True \
--max_template_date=2020-06-05

Primer z modulom za particijo CPU:

#!/bin/bash
#SBATCH -p cpu
#SBATCH -N 1
#SBATCH -c 8
#SBATCH -t 10:00:00

module load AlphaFold/2.2.2-foss-2021a-CUDA-11.3.1

alphafold \
--data_dir=/ceph/hpc/software/alphafold/ \
--fasta_paths=$(pwd)/sample.fasta \
--output_dir=$(pwd) \
--model_preset=multimer \
--db_preset=full_dbs \
--use_gpu_relax=False \
--max_template_date=2020-06-05

Dokumentacija