Clarification running ipd_design_pipeline_collab notebook

Thankyou for providing the example notebook of an end-to-end workflow!

I'm a little confused about the RF3 step. The first cell says that the first mpnn designed sequence is being folded, but it seems that the "atom_array" variable, still containing the coordinates of the original RFD3 backbone, is being provided as input:

# Create input from the MPNN-designed structure (first design)
# This re-folds the sequence to validate it adopts the intended structure
input_structure = InferenceInput.from_atom_array(atom_array, example_id="example_protein")
rf3_outputs = inference_engine.run(inputs=input_structure)

How should I correctly provide an mpnn-designed sequence instead? And can I predict the structure of all sequences as a batch?

Thanks, Ben

Start by opening the ipd_design_pipeline_collab notebook and reviewing the RF3 cell that creates InferenceInput from atom_array. Trace how the MPNN-designed sequence is represented and whether the notebook or pipeline documents batch prediction. Done means the notebook or its documentation clearly explains how to provide an MPNN sequence and predict multiple sequences.