How jinxedhuracan/frag-ml works
FRAG-ML is a computational drug-repurposing workflow that ranks candidate compounds against the ML2177c uridine nucleoside phosphorylase target. Its documented stages run from target/grid preparation through virtual screening, refined redocking, interaction and drug-likeness analysis, and fragment-informed rescoring, with a Streamlit interface for reviewing results. The sampled scripts support the interaction, ADMET, MCDA, FBDD, rescoring, complex-building, and visualization links; other stage ordering follows the README’s documented pipeline.
Main entry points: app.py Streamlit App · run_ml_rescorer.py ML Rescoring · PDB and PDBQT files Prepared Receptor · complexes_pdb Protein-Ligand Complexes
Components 22
Target Setup
- Target Structures
ML/01_structure - Prepared Receptor
ML/PDB and PDBQT files
Docking Workflow
- Virtual Screening
ML/run_screen_Final_attempt.py - Screen Results
ML/docking_results - Top-Hit Redocking
ML/scripts/redock_top15.py - Refined Poses
ML/redock_top15_results - Complex Builder
ML/build_complexes_mac.py - Protein-Ligand Complexes
ML/complexes_pdb
Candidate Analysis
- PLIP Profiling
ML/run_plip_analysis.py - Interaction Tables
ML/interaction_tables - ADMET Profiling
ML/run_admet_profiling.py - ADMET Profile
ML/admet_results
Ranking
- MCDA Ranking
ML/run_mcda_ranking.py - Repurposing Ranks
ML/final_submission_results - Fragment Metrics
ML/run_fragment_fbdd_analysis.py - FBDD Results
ML/fbdd_results - ML Rescoring
ML/run_ml_rescorer.py - Corrected Ranks
ML/ml_rescoring_results
Visualization
- Streamlit App
ML/app.py - 3D Views
ML/complexes_pdb
Other components
- Researcher
- Drug Library
Connections 31
- Target Structures → Prepared Receptor: prepares receptor
- Drug Library → Virtual Screening: screens compounds
- Prepared Receptor → Virtual Screening: docks against
- Virtual Screening → Screen Results: writes poses
- Screen Results → Top-Hit Redocking: selects hits
- Prepared Receptor → Top-Hit Redocking: redocks against
- Top-Hit Redocking → Refined Poses: writes refined poses
- Refined Poses → Complex Builder: assembles poses
- Prepared Receptor → Complex Builder: combines receptor
- Complex Builder → Protein-Ligand Complexes: writes complexes
- Protein-Ligand Complexes → PLIP Profiling: profiles interactions
- PLIP Profiling → Interaction Tables: writes interaction data
- Interaction Tables → MCDA Ranking: supplies contacts
- ADMET Profiling → ADMET Profile: writes descriptors
- ADMET Profile → MCDA Ranking: supplies drug properties
- MCDA Ranking → Repurposing Ranks: writes ranking
- Repurposing Ranks → Fragment Metrics: profiles ranked hits
- Fragment Metrics → FBDD Results: writes fragment metrics
- Repurposing Ranks → ML Rescoring: supplies candidates
- ADMET Profile → ML Rescoring: supplies descriptors
- FBDD Results → ML Rescoring: supplies efficiency metrics
- Interaction Tables → ML Rescoring: supplies contacts
- ML Rescoring → Corrected Ranks: writes corrected ranks
- Researcher → Streamlit App: reviews candidates
- Streamlit App → Repurposing Ranks: loads results
- Streamlit App → Corrected Ranks: loads results
- Streamlit App → ADMET Profile: loads profiles
- Streamlit App → FBDD Results: loads metrics
- Streamlit App → Interaction Tables: loads interactions
- Streamlit App → Protein-Ligand Complexes: loads complexes
- Protein-Ligand Complexes → 3D Views: provides structures