mirror of
https://github.com/RosettaCommons/RFdiffusion.git
synced 2026-06-04 18:44:21 +08:00
Implemented inpaint_str and added example
This commit is contained in:
@@ -25,6 +25,7 @@ inference:
|
||||
contigmap:
|
||||
contigs: null
|
||||
inpaint_seq: null
|
||||
inpaint_str: null
|
||||
provide_seq: null
|
||||
length: null
|
||||
|
||||
|
||||
12
examples/design_ppi_flexible_peptide.sh
Executable file
12
examples/design_ppi_flexible_peptide.sh
Executable file
@@ -0,0 +1,12 @@
|
||||
#!/bin/bash
|
||||
# Here, we're designing binders to the glp1 helical peptide, without specifying the topology of the binder a priori, and without specifying the structure of the peptide (we know peptides can be flexible).
|
||||
# We first provide the output path and input pdb of the target protein (5uul)
|
||||
# We then describe the protein we want with the contig input:
|
||||
# - residues 10-35 of the B chain of the target protein
|
||||
# - a chainbreak (as we don't want the binder fused to the target!)
|
||||
# - A 70-100 residue binder to be diffused (the exact length is sampled each iteration of diffusion)
|
||||
# We tell diffusion to target two specific residues on the target, specifically residues 28 and 29 of the B chain
|
||||
# We make 10 designs
|
||||
# We mask (diffuse) the structure of the peptide using the inpaint_str flag. This has the effect of having RFdiffusion simultaneously design a binder and predict the structure of the peptide within the complex.
|
||||
|
||||
../scripts/run_inference.py inference.output_prefix=example_outputs/design_ppi_flexible_peptide inference.input_pdb=input_pdbs/3IOL.pdb 'contigmap.contigs=[B10-35/0 70-100]' 'ppi.hotspot_res=[B28,B29]' inference.num_designs=10 'contigmap.inpaint_str=[B10-35]'
|
||||
1028
examples/input_pdbs/3IOL.pdb
Normal file
1028
examples/input_pdbs/3IOL.pdb
Normal file
File diff suppressed because it is too large
Load Diff
@@ -72,7 +72,7 @@ class Sampler:
|
||||
self.ckpt_path = conf.inference.ckpt_override_path
|
||||
print("WARNING: You're overriding the checkpoint path from the defaults. Check that the model you're providing can run with the inputs you're providing.")
|
||||
else:
|
||||
if conf.contigmap.inpaint_seq is not None or conf.contigmap.provide_seq is not None:
|
||||
if conf.contigmap.inpaint_seq is not None or conf.contigmap.provide_seq is not None or conf.contigmap.inpaint_str:
|
||||
# use model trained for inpaint_seq
|
||||
if conf.contigmap.provide_seq is not None:
|
||||
# this is only used for partial diffusion
|
||||
|
||||
Reference in New Issue
Block a user