How tiwarylab/alphafold2rave works
AlphaFold2-RAVE is a protein-ensemble workflow: it uses reduced-MSA AlphaFold2 predictions to generate structural diversity, learns collective variables with SPIB, then applies metadynamics to sample conformations and obtain Boltzmann-ranked results. The notebooks are documented as the workflow entry points; ravefuncs.py supplies clustering and OpenMM/PLUMED simulation helpers, while the lite module supports the no-simulation overview. Notebook internals were not sampled, so notebook-stage links reflect README documentation rather than inspected calls.
Main entry points: fullrunthrough_CSP.ipynb CSP run-through · rMSA_AF2_no_templates_no_relax.ipynb Reduced-MSA AF2 · structures.zip Diverse structures · ravefuncs.py Unbiased MD
Components 18
Structure prediction
- Reduced-MSA AF2
rMSA_AF2_no_templates_no_relax.ipynb - ColabFold AF2
- Diverse structures
CSP_data/structures.zip
CV learning
- Regular-space clustering
ravefuncs.py - SPIB CV learning
fullrunthrough_CSP.ipynb - SPIB configuration
sample_config.ini - Learned collective variables
CSP_data/COLVARS_AF2.dat
Enhanced sampling
- Bias input generation
ravefuncs.py - PDB repair
ravefuncs.py - Unbiased MD
ravefuncs.py - Metadynamics MD
ravefuncs.py - OpenMM and PLUMED
- PLUMED inputs
CSP_data/plumed_AF2.dat
Benchmark and results
- Sampling outputs
CSP_data/example_result - CSP run-through
fullrunthrough_CSP.ipynb - Light overview
ravefuncs_lite.py - CSP benchmark data
CSP_data
Other components
- Protein researcher
Connections 20
- Protein researcher → Reduced-MSA AF2: provides sequence
README.md - Reduced-MSA AF2 → ColabFold AF2: modifies workflow
README.md - Reduced-MSA AF2 → Diverse structures: generates structures
- Protein researcher → CSP run-through: runs protocol
README.md - Diverse structures → SPIB CV learning: learns CVs from
README.md - SPIB configuration → SPIB CV learning: configures
- SPIB CV learning → Learned collective variables: produces CVs
README.md - Diverse structures → Regular-space clustering: selects centers
- CSP run-through → Bias input generation: prepares bias
README.md - Learned collective variables → Bias input generation: defines bias coordinates
README.md - Bias input generation → PLUMED inputs: writes bias input
ravefuncs.py - CSP run-through → Unbiased MD: runs MD
README.md - Unbiased MD → PDB repair: repairs structure
ravefuncs.py - Unbiased MD → OpenMM and PLUMED: simulates with
ravefuncs.py - CSP run-through → Metadynamics MD: runs metadynamics
README.md - Metadynamics MD → OpenMM and PLUMED: simulates with
ravefuncs.py - Metadynamics MD → Sampling outputs: writes structures
- Sampling outputs → CSP run-through: ranks conformations
README.md - Protein researcher → Light overview: views overview
README.md - CSP benchmark data → CSP run-through: supports demo
README.md