View Code? Open in Web Editor NEW

PD-1 Targeted Antibody Discovery Using AI Protein Diffusion

Jupyter Notebook 26.64% TeX 1.19% Shell 44.82% Python 15.25% Makefile 0.29% Awk 3.15% Perl 0.48% C++ 7.26% Prolog 0.73% R 0.20%

pd-1_fab_diffusion's Introduction

PD-1 Targeted Antibody Discovery Using AI Protein Diffusion

Diffusion of Novel, PD1-targeting Fab Structures with in silico Evaluation

Generate Sequences using EvoDiff
- Follow EvoDiff logic in scripts/0_sequence_generation/EvoDiff_sequence_generation.md
Generate Structures using AlphaFold2 (TUPPD1-001 to TUPPD1-009)
- Run AlphaFold2 Batch using scripts/1_structure_generation/AlphaFold2_batch.ipynb
Prepare Structures
- Renumber L chains in PDB files with PyMol: scripts/2_structure_prep/PDB_renumberer.ipynb
- Detect CDR loop residues: scripts/2_structure_prep/fab_cdr_detection.ipynb
Generate HADDOCK experiment files
- Run experiment generation using scripts/3_experiment_generation/make_HADDOCK_experiments.ipynb
Submit HADDOCK experiments to HPC
- Create Singularity container from Docker image: singularity build haddock.sif docker://cford38/haddock:2.4_36cores
- Run experiment generation: bash ./scripts/4_experiment_submission/submit_experiments.sh
  - (Remember to chmod -R 755 the experiments/ folder so that Singularity can execute the scripts.)
Collect best docked PDB structure and metrics for each experiment
- Run experiment metrics collection: scripts/5_metrics_collection/postprocessing.ipynb
  - This notebook also runs PRODIGY to calculate binding affinities for each structure.
Clean up HADDOCK run files
- Run cleanup to move each experiments' run1/ directory into outputs: scripts/6_cleanup/cleanup.ipynb

Recommend Projects