What input does the design workflow need?
Provide a protein structure, either by uploading a PDB file or by entering a PDB ID. The system can then discover a pocket automatically or use a prepared pocket file.
Which parameters matter most?
Ligand size controls the scale of generated molecules, num_molecules controls how many candidates are sampled, and seed makes a run reproducible.
What is the output?
Each job stores the extracted pocket, generated PDB files for the designed molecules, validation logs, and a summary JSON file.
What checkpoint is used?
The default model is the bundled MOAD checkpoint at /home/html/share/software/medusadock/checkpoints/moad.ckpt unless a different checkpoint is supplied in the job payload.
Why did my run fail?
Common causes are a malformed protein file, an unselectable pocket, or the design worker failing validation for all sampled molecules. Check the per-molecule stderr and validation logs.