Strategies for retaining the hydrogen-bond donor/acceptor architecture defined in the RFD3 active-site model during LigandMPNN sequence optimization

Dear RFD3 team,

Thank you for your excellent work on RFdiffusion3. I have a question regarding enzyme design and the preservation of catalytic hydrogen-bond interactions throughout the design pipeline.

If I define specific hydrogen-bond donor and acceptor atoms between my substrate and catalytic residues in the initial catalytic model, how can I determine whether the generated RFD3 scaffolds actually support and preserve these intended hydrogen-bond interactions?

In addition, after running LigandMPNN, how can I ensure that residues forming these substrate hydrogen bonds are retained and not redesigned away? Is it recommended to fix these residues during sequence design, or is there another preferred strategy?

Thank you for your help.

No files, tests, or entry points are named. Start by reviewing the RFD3 scaffold-generation and LigandMPNN sequence-design entry points to determine how catalytic hydrogen-bond constraints are represented. Done would require a project-supported strategy for checking preservation during scaffold generation and preventing or handling redesign during sequence optimization.