AFToolkit is a framework for molecular modeling of proteins with AlphaFold2 derived representations.
git clone https://github.com/AIRI-Institute/AFToolkit.git
cd AFToolkit
conda env create --name=aftoolkit_env -f environment.yml
source activate aftoolkit_env
pip install .
We prepared 3 models for protein stability task usinf SVM, MLP and Catboost adapters, and 1 model for prediction protein-protein affinity using SVM adapter.
You can download models using foolowing links:
- Models trained on protein stability task:
mkdir model/monomer
mkdir model/monomer/pair+lddt_logits+plddt/
cd model/monomer/pair+lddt_logits+plddt/
wget -O https://bioinformatics-kardymon.obs.ru-moscow-1.hc.sbercloud.ru/AFToolkit/models/monomer/trained_catboost_concat_nomultitrain_aggmutpos_multisum.pkl
wget -O https://bioinformatics-kardymon.obs.ru-moscow-1.hc.sbercloud.ru/AFToolkit/models/monomer/trained_mlp_concat_nomultitrain_aggmutpos_multisum.pkl
wget -O https://bioinformatics-kardymon.obs.ru-moscow-1.hc.sbercloud.ru/AFToolkit/models/monomer/trained_svm_concat_nomultitrain_aggmutpos_multisum.pkl
- Model trained on protein-protein affinity task:
mkdir model/multimer
mkdir model/multimer/pair+lddt_logits+plddt/
cd model/multimer/pair+lddt_logits+plddt/
wget -O https://bioinformatics-kardymon.obs.ru-moscow-1.hc.sbercloud.ru/AFToolkit/models//multimer/trained_svm_concat_nomultitrain_aggmutpos_multisum.pkl
To test our models, please, download task csv-files and features pkl-files. File with features for training monomer adapter models are too large, please, generate them using run_protein_task scripts and cdna+PROSTATA_mut_idxs.csv task file.
- Test data for protein stability task:
cd data
#task files
wget -O https://bioinformatics-kardymon.obs.ru-moscow-1.hc.sbercloud.ru/AFToolkit/data/stability_task_files.zip
#s669 features pkl-files
wget -O https://bioinformatics-kardymon.obs.ru-moscow-1.hc.sbercloud.ru/AFToolkit/data/s699_pkls.zip
#ssym features pkl-files
wget -O https://bioinformatics-kardymon.obs.ru-moscow-1.hc.sbercloud.ru/AFToolkit/data/ssym_pkls.zip
#PTmul features pkl-files
wget -O https://bioinformatics-kardymon.obs.ru-moscow-1.hc.sbercloud.ru/AFToolkit/data/protherm_pkls.zip
#cDNA de novo features pkl-files
wget -O https://bioinformatics-kardymon.obs.ru-moscow-1.hc.sbercloud.ru/AFToolkit/data/denovo_pkls.zip
#cDNA indels features pkl-files
wget -O https://bioinformatics-kardymon.obs.ru-moscow-1.hc.sbercloud.ru/AFToolkit/data/cdna_indel_pkls.zip
- Test data for protein-protein affinity task:
cd data
#task files
wget -O https://bioinformatics-kardymon.obs.ru-moscow-1.hc.sbercloud.ru/AFToolkit/data/ppi_task_files.zip
#S4196 features pkl-files
wget -O https://bioinformatics-kardymon.obs.ru-moscow-1.hc.sbercloud.ru/AFToolkit/data/S4169_pkl.zip
#C380 features pkl-files
wget -O https://bioinformatics-kardymon.obs.ru-moscow-1.hc.sbercloud.ru/AFToolkit/data/C380_pkl.zip
You can calculate AF2 features for certain protein or protein complex:
run_protein_task \
--pdb data/examples/PDB/1A7V.pdb \
--chain A \
--mutations A:A66H \
-n 3 \
--return-all-cycles \
-o ./output
run_protein_complex_task \
--pdb data/examples/PDB/1CSE.pdb \
--chains-for-protein I \
--mutations I:L38S \
-n 1 \
--return-all-cycles \
-o ./output
You can calculate AF2 features for the dataset using task-file. Please, see examples.
run_protein_task \
-i data/examples/monomer_task.csv \
--output-dir ./output \
-n 3
run_protein_complex_task \
-i data/examples/ppi_task.csv \
--output-dir ./output \
-n 3
To see description for all optional arguments of run_protein_task and run_protein_complex_task scripts use help option.
run_protein_task -h
run_protein_complex_task -h
You can also using implemented functions inside your python code and considered protein embeddings for other tasks or save recycled protein structure in PDB file.
from AFToolKit.processing.protein_task import ProteinTask
from AFToolKit.processing.openfold_wrapper import OpenFoldWrapper
from AFToolKit.processing.arg_parser import parse_mutations
from AFToolKit.processing.utils import save_to_pdb
of_wrapper = OpenFoldWrapper(device='cuda:0',
inference_n_recycle=3,
always_use_template=False,
side_chain_mask=False,
return_all_cycles=False)
of_wrapper.init_model()
protein_task = ProteinTask()
protein_task.set_input_protein_task(protein_path='data/examples/PDB/1A0F.pdb',
chains=['A'])
protein_task.set_task_mutants(parse_mutations('A:S11A'))
protein_task.set_observable_positions()
logger.info('Calculate embeddings')
protein_task.evaluate(of_wrapper=of_wrapper,
store_of_protein=args.store_protein
)
#save recycled wildtype structure
save_to_pdb(protein_task.get_wildtype_protein_of(), 'data/examples/test_save.pdb')
You can evaluate our pretrained SVM, MLP and CatBoost models.
Please see examples for protein stability task in run_test_monomer_example.py file.
For affinity prediction use run_test_multimer_example.py file and for rosetta energy prediction use run_test_perresidue_example.py.
You can create your own dataset using run_protein_task and run_protein_complex_task scripts and train the corresponding model or create your own adapter.
Please see examples in train_{adaptor}_monomer.py,train_svm_multimer.py and train_svm_perresidue files.
This code is provided under MIT License:
The MIT License (MIT) Copyright (c) 2016 AYLIEN Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.