AF3Score Input PDB files

[adxyzz]

I have a naive question to ask: Where do the input PDB files in AF3Score come from?
Are the input PDB files in the AF3Score from the PDB files generated by the PPIFlow step, or from the PDB files obtained by predicting all the newly generated sequences with AF3 after ProteinMPNN generates the sequences? Thank you very much.

[MacroHongZ]

AF3Score input pdb is the structure where FlowPacker filled the side chains of the MPNN design sequence. Please refer to https://github.com/Mingchenchen/PPIFlow/blob/main/demo_vhh.ipynb

### Step.1 Backbone Generation & Sequence Design

- **Backbone**: Sampling diverse CDR configurations that fit the target epitope using **PPIFlow**.
- **Sequence**: Inverse folding and amino acid sequence optimization using **AbMPNN** .

### Step.2 Scoring & Filtering

- **Validation**: High-throughput screening of designs using **FlowPacker** and **AF3Score**.