rfdiffusion

Name: rfdiffusion
Author: adaptyvbio

31installs20forks

How to install rfdiffusion?

adaptyvbio/protein-design-skills/rfdiffusion

$npx -y skills add adaptyvbio/protein-design-skills --skill rfdiffusion

Installs into the current project.

Summary

Generate protein backbones using RFdiffusion, a diffusion-based generative model for de novo protein structure generation. Use this skill when: (1) Designing binder scaffolds for a target protein, (2) Generating novel protein backbones from scratch, (3) Scaffolding functional mot

Files · 1

View on GitHub

SKILL.md

1	# RFdiffusion Backbone Generation
2
3	## Prerequisites
4
5	\| Requirement \| Minimum \| Recommended \|
6	\|-------------\|---------\|-------------\|
7	\| Python \| 3.9+ \| 3.10 \|
8	\| CUDA \| 11.7+ \| 12.0+ \|
9	\| GPU VRAM \| 16GB \| 24GB (A10G) \|
10	\| RAM \| 16GB \| 32GB \|
11
12	## How to run
13
14	RFdiffusion is not in biomodals, so run it from the official RosettaCommons repo or
15	its Docker image, not through Modal.
16
17	### Local installation (official repo)
18	```bash
19	git clone https://github.com/RosettaCommons/RFdiffusion.git
20	cd RFdiffusion
21
22	# Conda env including the required NVIDIA SE(3)-Transformer
23	conda env create -f env/SE3nv.yml
24	conda activate SE3nv
25	cd env/SE3Transformer && pip install . && cd ../..
26	pip install -e .
27
28	# Download weights (per-file hashed paths; see the repo README for the full list)
29	mkdir -p models
30	wget -P models http://files.ipd.uw.edu/pub/RFdiffusion/e29311f6f1bf1af907f9ef9f44b8328b/Complex_base_ckpt.pt
31
32	# Binder design run; single-quote the hydra args so the shell does not split [] or ,
33	./scripts/run_inference.py \
34	inference.input_pdb=target.pdb \
35	'contigmap.contigs=[A1-150/0 70-100]' \
36	'ppi.hotspot_res=[A45,A67,A89]' \
37	inference.num_designs=100
38	```
39
40	A RosettaCommons-maintained Docker image is also available from the repo README.
41	After backbone generation, design sequences with `proteinmpnn`.
42
43	## Config Schema (Hydra)
44
45	### Contigmap Syntax
46	```bash
47	# De novo single chain (50-100 residues)
48	contigmap.contigs=[50-100]
49
50	# Binder + target (A = target chain, fixed with /0)
51	contigmap.contigs=[A1-150/0 70-100]
52
53	# Motif scaffolding (preserve residues, /0 = fixed)
54	contigmap.contigs=[20-40/0 A10-30/0 20-40]
55
56	# Multi-chain binder
57	contigmap.contigs=[A1-100/0 B1-100/0 60-80]
58
59	# Variable length ranges
60	contigmap.contigs=[A1-150/0 50-100] # Binder 50-100 AA
61	```
62
63	### Hotspot Specification
64	```bash
65	# Residues for interface (chain + resnum, no spaces)
66	ppi.hotspot_res=[A45,A67,A89]
67	```
68
69	## Common mistakes
70
71	### Contig Syntax
72	✅ Correct:
73	```bash
74	'contigmap.contigs=[A1-150/0 70-100]' # Target fixed (/0), binder variable
75	```
76
77	Single-quote the whole argument so the shell does not split on the space inside the
78	brackets.
79
80	❌ Wrong:
81	```bash
82	contigmap.contigs=[A1-150 70-100] # Missing /0 - target will move!
83	contigmap.contigs=[A1-150/0 70-100] # Unquoted: shell splits on the space
84	contigmap.contigs=[A1-150/0, 70-100] # Extra comma changes the contig string
85	```
86
87	### Hotspot Residues
88	✅ Correct:
89	```bash
90	'ppi.hotspot_res=[A45,A67,A89]' # Chain letter + residue number, whole arg quoted
91	```
92
93	❌ Wrong:
94	```bash
95	ppi.hotspot_res=[45,67,89] # Missing chain letter
96	'ppi.hotspot_res=[A45, A67, A89]' # Spaces inside the list break parsing
97	```
98
99	### Complete Parameter Reference
100
101	#### Core Parameters
102	\| Parameter \| Default \| Range \| Description \|
103	\|-----------\|---------\|-------\|-------------\|
104	\| `inference.num_designs` \| 10 \| 1-10000 \| Number of designs to generate \|
105	\| `inference.input_pdb` \| - \| path \| Target structure file \|
106	\| `inference.output_prefix` \| output \| string \| Output filename prefix \|
107	\| `diffuser.T` \| 50 \| 20-200 \| Diffusion timesteps \|
108	\| `denoiser.noise_scale_ca` \| 1.0 \| 0.0-2.0 \| CA atom noise (0.5-0.8 = conservative) \|
109	\| `denoiser.noise_scale_frame` \| 1.0 \| 0.0-2.0 \| Frame noise \|
110	\| `inference.ckpt_override_path` \| - \| path \| Model checkpoint \|
111	\| `potentials.guide_scale` \| 1.0 \| 0.1-10 \| Guidance strength \|
112	\| `potentials.guide_decay` \| constant \| string \| Decay type \|
113
114	#### Advanced Parameters
115	\| Parameter \| Default \| Description \|
116	\|-----------\|---------\|-------------\|
117	\| `diffuser.partial_T` \| None \| Start diffusion from timestep T (partial diffusion) \|
118	\| `contigmap.inpaint_str` \| None \| Sequence positions to inpaint \|
119	\| `scaffoldguided.scaffoldguided` \| false \| Enable scaffold-guided generation \|
120	\| `scaffoldguided.target_pdb` \| None \| Scaffold template PDB \|
121	\| `ppi.binderlen` \| None \| Specify exact binder length \|
122
123	#### Symmetry Parameters
124	\| Parameter \| Default \| Description \|
125	\|-----------\|---------\|-------------\|
126	\| `symmetry.symmetry` \| None \| Symmetry type (C2, C3, C4, D2, etc.) \|
127	\| `symmetry.recenter` \| true \| Recenter symmetric assembly \|
128	\| `symmetry.radius` \| None \| Radius constraint for symmetric assembly \|
129
130	#### Fold Conditioning
131	\| Paramet

People who liked this also viewed

6 picks

Type

rfdiffusion

31installs20forks

View on GitHub

How to install rfdiffusion?

adaptyvbio/protein-design-skills/rfdiffusion

$npx -y skills add adaptyvbio/protein-design-skills --skill rfdiffusion

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# RFdiffusion Backbone Generation
2
3	## Prerequisites
4
5	\| Requirement \| Minimum \| Recommended \|
6	\|-------------\|---------\|-------------\|
7	\| Python \| 3.9+ \| 3.10 \|
8	\| CUDA \| 11.7+ \| 12.0+ \|
9	\| GPU VRAM \| 16GB \| 24GB (A10G) \|
10	\| RAM \| 16GB \| 32GB \|
11
12	## How to run
13
14	RFdiffusion is not in biomodals, so run it from the official RosettaCommons repo or
15	its Docker image, not through Modal.
16
17	### Local installation (official repo)
18	```bash
19	git clone https://github.com/RosettaCommons/RFdiffusion.git
20	cd RFdiffusion
21
22	# Conda env including the required NVIDIA SE(3)-Transformer
23	conda env create -f env/SE3nv.yml
24	conda activate SE3nv
25	cd env/SE3Transformer && pip install . && cd ../..
26	pip install -e .
27
28	# Download weights (per-file hashed paths; see the repo README for the full list)
29	mkdir -p models
30	wget -P models http://files.ipd.uw.edu/pub/RFdiffusion/e29311f6f1bf1af907f9ef9f44b8328b/Complex_base_ckpt.pt
31
32	# Binder design run; single-quote the hydra args so the shell does not split [] or ,
33	./scripts/run_inference.py \
34	inference.input_pdb=target.pdb \
35	'contigmap.contigs=[A1-150/0 70-100]' \
36	'ppi.hotspot_res=[A45,A67,A89]' \
37	inference.num_designs=100
38	```
39
40	A RosettaCommons-maintained Docker image is also available from the repo README.
41	After backbone generation, design sequences with `proteinmpnn`.
42
43	## Config Schema (Hydra)
44
45	### Contigmap Syntax
46	```bash
47	# De novo single chain (50-100 residues)
48	contigmap.contigs=[50-100]
49
50	# Binder + target (A = target chain, fixed with /0)
51	contigmap.contigs=[A1-150/0 70-100]
52
53	# Motif scaffolding (preserve residues, /0 = fixed)
54	contigmap.contigs=[20-40/0 A10-30/0 20-40]
55
56	# Multi-chain binder
57	contigmap.contigs=[A1-100/0 B1-100/0 60-80]
58
59	# Variable length ranges
60	contigmap.contigs=[A1-150/0 50-100] # Binder 50-100 AA
61	```
62
63	### Hotspot Specification
64	```bash
65	# Residues for interface (chain + resnum, no spaces)
66	ppi.hotspot_res=[A45,A67,A89]
67	```
68
69	## Common mistakes
70
71	### Contig Syntax
72	✅ Correct:
73	```bash
74	'contigmap.contigs=[A1-150/0 70-100]' # Target fixed (/0), binder variable
75	```
76
77	Single-quote the whole argument so the shell does not split on the space inside the
78	brackets.
79
80	❌ Wrong:
81	```bash
82	contigmap.contigs=[A1-150 70-100] # Missing /0 - target will move!
83	contigmap.contigs=[A1-150/0 70-100] # Unquoted: shell splits on the space
84	contigmap.contigs=[A1-150/0, 70-100] # Extra comma changes the contig string
85	```
86
87	### Hotspot Residues
88	✅ Correct:
89	```bash
90	'ppi.hotspot_res=[A45,A67,A89]' # Chain letter + residue number, whole arg quoted
91	```
92
93	❌ Wrong:
94	```bash
95	ppi.hotspot_res=[45,67,89] # Missing chain letter
96	'ppi.hotspot_res=[A45, A67, A89]' # Spaces inside the list break parsing
97	```
98
99	### Complete Parameter Reference
100
101	#### Core Parameters
102	\| Parameter \| Default \| Range \| Description \|
103	\|-----------\|---------\|-------\|-------------\|
104	\| `inference.num_designs` \| 10 \| 1-10000 \| Number of designs to generate \|
105	\| `inference.input_pdb` \| - \| path \| Target structure file \|
106	\| `inference.output_prefix` \| output \| string \| Output filename prefix \|
107	\| `diffuser.T` \| 50 \| 20-200 \| Diffusion timesteps \|
108	\| `denoiser.noise_scale_ca` \| 1.0 \| 0.0-2.0 \| CA atom noise (0.5-0.8 = conservative) \|
109	\| `denoiser.noise_scale_frame` \| 1.0 \| 0.0-2.0 \| Frame noise \|
110	\| `inference.ckpt_override_path` \| - \| path \| Model checkpoint \|
111	\| `potentials.guide_scale` \| 1.0 \| 0.1-10 \| Guidance strength \|
112	\| `potentials.guide_decay` \| constant \| string \| Decay type \|
113
114	#### Advanced Parameters
115	\| Parameter \| Default \| Description \|
116	\|-----------\|---------\|-------------\|
117	\| `diffuser.partial_T` \| None \| Start diffusion from timestep T (partial diffusion) \|
118	\| `contigmap.inpaint_str` \| None \| Sequence positions to inpaint \|
119	\| `scaffoldguided.scaffoldguided` \| false \| Enable scaffold-guided generation \|
120	\| `scaffoldguided.target_pdb` \| None \| Scaffold template PDB \|
121	\| `ppi.binderlen` \| None \| Specify exact binder length \|
122
123	#### Symmetry Parameters
124	\| Parameter \| Default \| Description \|
125	\|-----------\|---------\|-------------\|
126	\| `symmetry.symmetry` \| None \| Symmetry type (C2, C3, C4, D2, etc.) \|
127	\| `symmetry.recenter` \| true \| Recenter symmetric assembly \|
128	\| `symmetry.radius` \| None \| Radius constraint for symmetric assembly \|
129
130	#### Fold Conditioning
131	\| Paramet

People who liked this also viewed

6 picks

Type