RFdiffusion icon indicating copy to clipboard operation
RFdiffusion copied to clipboard

partial diffusion with an antibody-antigen complex structure

Open 957-z opened this issue 5 months ago • 1 comments

I have an antigen-antibody complex structure, and I performed diffusion on part of the antibody sequence. The current parameters are as follows, but the result connects the two chains of the antibody together. How should I modify the parameters?

run_inference.py inference.output_prefix=rfd_diffuser20/test inference.input_pdb=test.pdb 'contigmap.contigs=["228-228/0 C229-503"]' diffuser.partial_T=20 inference.num_designs=1000 'contigmap.provide_seq=[0-29,31-227]' 'ppi.hotspot_res=[C262,C263,C265,C266,C267,C268,C269,C270,C315,C316,C317,C319,C321,C329,C344,C345,C346,C348,C374,C375,C376,C377,C393,C395,C397,C399,C434,C435,C436,C437,C438,C451,C455,C457,C458,C459,C460,C462,C465,C467,C469]'

957-z avatar Aug 17 '25 02:08 957-z

Your contig map is calling for two chains, only one of which is generated from scratch. One of length 228, created from scratch, and one which is the input PDB residues 229-503 of chain C.

If you want RFdiffusion to generate two chains, you need to tell it to add a chainbreak in the proper spot. For example, if you're looking for a heavy and light chain both of length 114, then you would want something like:

'contigmap.contigs=["114-114/0 114-114/0 C229-503"]'

roccomoretti avatar Aug 18 '25 14:08 roccomoretti