Protein Models: Autoimmunity-associated DIORA1 binds the MRCK family of serine/threonine kinases and controls cell motility

Models generated for the interaction between DIORA1 and MRCKB with massive sampling using AFsample (Wallner, 2022) File descriptions The data is compressed with zstandard to extract an archive use the following command: tar --use-compress-program=unzstd -xvf all_models_scores.tar.zst all_models_scores.tar.zst - Contains scores for all generated models, extracts to the folder all_models/selected_models.tar.zst - Contains top-ranked selected models that are also highlighted in the paper, extracts to selected_models/ Models for different interaction partners All models and corresponding results *.pkl (distogram removed to save space) are archived in different files for each interaction partner. Stoichiometry 1:1 (main text) KIM_diora_1_1.tar.zst (2.2GB) - All 3000 models generated for the KIM region of MRCKB to interact with DIORA1.MRCKB_1020_1582-diora1_1_214.tar.zst (3.9GB) - All 3,000 models generated for the MRCKB range 1020-1582 corresponding to the C1-PH-CNH domains. Different models with different stoichiometries (Supplementary Information) KIM_diora1_1_2.tar.zst (1.9GB) - 3,000 models for KIM:DIORA1 1:2KIM_diora1_2_1.tar.zst (1.4GB) - 3,000 models for KIM:DIORA1 2:1KIM_diora1_2_2.tar.zst (2.6GB) - 3,000 models for KIM:DIORA1 2:2 MRCKB_1020_1582-diora1_1_214_1_2.tar.zst (5.3GB) - 3,000 models for C1-PH-CNH:DIORA1 1:2MRCKB_1020_1582-diora1_1_214_2_1.tar.zst (8.8GB) - 3,000 models for C1-PH-CNH:DIORA1 2:1MRCKB_1020_1582-diora1_1_214_2_2.tar.zst (10.8GB) - 3,000 models for C1-PH-CNH:DIORA1 2:2 Modelling protocol Models generated for the interaction between DIORA1 and MRCKB with massive sampling using AFsample (Wallner, 2022) Code: http://wallnerlab.org/AFsample Alphafold was run with dropout activated at inference using all 15 (3x5) neural network models available for multimeric prediction: model_[1-5]_multimer model_[1-5]_multimer_v2 model_[1-5]_multimer_v3 The following command was used to generate 200 models per neural network model, 3000 (15x200) models in total for each modeling input python alphafoldv2.3.1/run_alphafold.py--dropout--nstruct 200--use_precomputed_msas --data_dir=/proj/beyondfold/apps/alphafold_data --template_mmcif_dir=/proj/beyondfold/apps/alphafold_data/pdb_mmcif/mmcif_files --obsolete_pdbs_path=/proj/beyondfold/apps/alphafold_data/pdb_mmcif/obsolete.dat --mgnify_database_path=/proj/beyondfold/users/x_bjowa/databases/mgnify/mgy_all.fa --uniref90_database_path=/proj/beyondfold/apps/alphafold_data/uniref90/uniref90.fasta --max_template_date=3000-01-01 --use_gpu_relax=True --model_preset=multimer --db_preset=full_dbs --uniref30_database_path=/proj/beyondfold/apps/alphafold_data/uniclust30/UniRef30_2021_03/UniRef30_2021_03 --bfd_database_path=/proj/beyondfold/apps/alphafold_data/bfd/bfd_metaclust_clu_complete_id30_c90_final_seq.sorted_opt --pdb_seqres_database_path=/proj/beyondfold/apps/alphafold_data/pdb_seqres/pdb_seqres.txt --uniprot_database_path=/proj/beyondfold/apps/alphafold_data/uniprot/uniprot.fasta本数据集为利用AFsample（Wallner, 2022）开展大规模采样所生成的DIORA1与MRCKB相互作用模型。 ## 文件说明 本数据集采用zstandard压缩，解压归档文件需使用如下命令： tar --use-compress-program=unzstd -xvf all_models_scores.tar.zst all_models_scores.tar.zst：包含所有生成模型的评分，解压后将生成all_models/文件夹；selected_models.tar.zst：包含本文献重点标注的高分精选模型，解压后将生成selected_models/文件夹。 ## 不同相互作用靶点对应的模型 所有模型及对应结果（为节省存储空间已移除距离图（distogram））均按不同相互作用靶点归档为独立文件，文件格式为*.pkl。 ### 化学计量比1:1（正文章节） KIM_diora_1_1.tar.zst（2.2GB）：包含MRCKB的KIM区域与DIORA1相互作用的全部3000个模型。 MRCKB_1020_1582-diora1_1_214.tar.zst（3.9GB）：包含对应C1-PH-CNH结构域的MRCKB 1020-1582区域与DIORA1相互作用的全部3000个模型。 ### 不同化学计量比的模型（补充材料章节） KIM_diora1_1_2.tar.zst（1.9GB）：包含KIM与DIORA1以1:2化学计量比相互作用的3000个模型。 KIM_diora1_2_1.tar.zst（1.4GB）：包含KIM与DIORA1以2:1化学计量比相互作用的3000个模型。 KIM_diora1_2_2.tar.zst（2.6GB）：包含KIM与DIORA1以2:2化学计量比相互作用的3000个模型。 MRCKB_1020_1582-diora1_1_214_1_2.tar.zst（5.3GB）：包含C1-PH-CNH结构域与DIORA1以1:2化学计量比相互作用的3000个模型。 MRCKB_1020_1582-diora1_1_214_2_1.tar.zst（8.8GB）：包含C1-PH-CNH结构域与DIORA1以2:1化学计量比相互作用的3000个模型。 MRCKB_1020_1582-diora1_1_214_2_2.tar.zst（10.8GB）：包含C1-PH-CNH结构域与DIORA1以2:2化学计量比相互作用的3000个模型。 ## 建模流程 本数据集为利用AFsample（Wallner, 2022）生成的DIORA1与MRCKB相互作用模型，具体建模流程如下： 代码地址：http://wallnerlab.org/AFsample 本研究在推理阶段启用Dropout，采用全部15个（3组×5个）可用的多聚体预测神经网络模型： model_[1-5]_multimer、model_[1-5]_multimer_v2及model_[1-5]_multimer_v3 针对每个建模输入，每个神经网络模型生成200个模型，总计生成3000个（15×200）模型，所用命令如下： python alphafoldv2.3.1/run_alphafold.py --dropout --nstruct 200 --use_precomputed_msas --data_dir=/proj/beyondfold/apps/alphafold_data --template_mmcif_dir=/proj/beyondfold/apps/alphafold_data/pdb_mmcif/mmcif_files --obsolete_pdbs_path=/proj/beyondfold/apps/alphafold_data/pdb_mmcif/obsolete.dat --mgnify_database_path=/proj/beyondfold/users/x_bjowa/databases/mgnify/mgy_all.fa --uniref90_database_path=/proj/beyondfold/apps/alphafold_data/uniref90/uniref90.fasta --max_template_date=3000-01-01 --use_gpu_relax=True --model_preset=multimer --db_preset=full_dbs --uniref30_database_path=/proj/beyondfold/apps/alphafold_data/uniclust30/UniRef30_2021_03/UniRef30_2021_03 --bfd_database_path=/proj/beyondfold/apps/alphafold_data/bfd/bfd_metaclust_clu_complete_id30_c90_final_seq.sorted_opt --pdb_seqres_database_path=/proj/beyondfold/apps/alphafold_data/pdb_seqres/pdb_seqres.txt --uniprot_database_path=/proj/beyondfold/apps/alphafold_data/uniprot/uniprot.fasta