Generating a 'dual' protein binder

[KodyKlupt]

Looking for some advice on generating a binder that links protein 1 (chain A) and protein 2 (chain B), separated by a fixed distance in coordinate space. Perhaps can also consider this as a fusion protein (though not one sequence chain) where the middle protein is the de novo binder to both terminals.

Wondering what, if possible, the contigs would look like? Given a PDB file with chain A and B can RFDiffusion support e.g. A1-100/0 B1-100/0 50-100 (both proteins 100 residues, with a 50-100 a.a. binder)?

Thanks!

[roccomoretti]

RFdiffusion works on a "residue gas" basis, with chains more-or-less functioning purely on connectivity annotation basis. As such, it should be robust to having the known context being provided as multiple chains.

Your contig specification should work as you expect it to. I will note, however, that there's nothing in the contig specification which requires the generated protein to be built between chains A & B. You probably want to provide hotspot residues on both chains to encourage RFdiffusion to generate backbones which contact both chains.