Examples

• Pinder index
  • Download the dataset
  • Pinder metadata API
  • Pinder index API
• Pinder system
  • PinderSystem API - base class representing Structure’s in a pinder entry
  • Classify system difficulty based on degree of conformational shift in unbound and bound
  • Illustrating utilities available in Structure instances
  • Can also filter “in place” rather than returning a copy (a la pandas)
  • Create masked unbound complex aligned to bound for apo
  • Structures have resolved_pdb2uniprot and resolved_uniprot2pdb properties
  • Case with multiple available apo structures
  • Classify system difficulty based on degree of conformational shift in unbound and bound
  • A bunch of other features of the Structure class are illustrated below
• Pinder loader
  • Accessing and loading data for training
  • Constructing torch datasets and dataloaders from pinder systems
  • Implementing your own PyTorch Dataset & DataLoader for pinder
• Pinder eval
  • Harnessing Biotite: the workhorse beind Pinder’s abstractions
  • Evaluation harness: implementation of DockQ metrics
  • pinder_eval entrypoint
  • References
• FAQ: cropped monomer superposition
  • Examine monomer shapes prior to cropping or superposition
  • Single alternative monomer use-case without cropping
  • With cropping such that apo and holo have the same shapes
  • Why are these atoms missing in the apo, but present in the holo structure?
  • What if we need them to be identical shapes?
  • We specified monomer_types apo, but got a tuple of structures that included the predicted complex, why?
• MLSB PINDER Challenge
  • Rules for valid model training
  • Rules for valid inference submissions
  • Evaluation dataset
  • Accessing and loading data for training