proteinmpnn

Name: proteinmpnn
Author: adaptyvbio

44installs20forks

How to install proteinmpnn?

adaptyvbio/protein-design-skills/proteinmpnn

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

Installs into the current project.

Summary

Design protein sequences using ProteinMPNN inverse folding. Use this skill when: (1) Designing sequences for RFdiffusion backbones, (2) Redesigning existing protein sequences, (3) Fixing specific residues while designing others, (4) Optimizing sequences for expression or stabilit

Files · 1

View on GitHub

SKILL.md

1	# ProteinMPNN Sequence Design
2
3	## Prerequisites
4
5	\| Requirement \| Minimum \| Recommended \|
6	\|-------------\|---------\|-------------\|
7	\| Python \| 3.8+ \| 3.10 \|
8	\| CUDA \| 11.0+ \| 11.7+ \|
9	\| GPU VRAM \| 8GB \| 16GB (T4) \|
10	\| RAM \| 8GB \| 16GB \|
11
12	## How to run
13
14	> First time? See [Getting started](../../docs/getting-started.md) to set up Modal and biomodals.
15
16	### Option 1: Local installation (recommended)
17	```bash
18	git clone https://github.com/dauparas/ProteinMPNN.git
19	cd ProteinMPNN
20
21	python protein_mpnn_run.py \
22	--pdb_path backbone.pdb \
23	--out_folder output/ \
24	--num_seq_per_target 16 \
25	--sampling_temp "0.1"
26	```
27
28	GPU: T4 (16GB) sufficient \| Time: ~50-100 sequences/minute
29
30	### Option 2: Modal (via LigandMPNN wrapper)
31	```bash
32	cd biomodals
33	# modal_ligandmpnn.py takes --input-pdb and forwards run.py args via --params-str
34	modal run modal_ligandmpnn.py \
35	--input-pdb backbone.pdb \
36	--params-str "--model_type protein_mpnn --number_of_batches 16 --temperature 0.1"
37	```
38
39	GPU (Modal): A10G default \| Timeout: 900s default
40
41	Note: LigandMPNN includes ProteinMPNN functionality (select with `--model_type protein_mpnn`).
42
43	## Config Schema
44
45	### Core Parameters
46
47	\| Parameter \| Default \| Range \| Description \|
48	\|-----------\|---------\|-------\|-------------\|
49	\| `--pdb_path` \| required \| path \| Single PDB input \|
50	\| `--pdb_path_chains` \| all \| A,B \| Chains to design (comma-sep) \|
51	\| `--out_folder` \| required \| path \| Output directory \|
52	\| `--num_seq_per_target` \| 1 \| 1-1000 \| Sequences per structure \|
53	\| `--sampling_temp` \| "0.1" \| "0.0001-1.0" \| Temperature (string!) \|
54	\| `--seed` \| 0 \| int \| Random seed \|
55	\| `--batch_size` \| 1 \| 1-32 \| Batch size \|
56
57	### Temperature Guide
58	```
59	0.1 -> Low diversity, high recovery (production)
60	0.2 -> Moderate diversity (default)
61	0.3 -> Higher diversity (exploration)
62	0.5+ -> Very diverse, lower quality
63	```
64
65	IMPORTANT: Temperature must be passed as a string, not float.
66
67	## Common mistakes
68
69	### Temperature Parameter
70	✅ Correct:
71	```bash
72	--sampling_temp "0.1" # String with quotes
73	```
74
75	❌ Wrong:
76	```bash
77	--sampling_temp 0.1 # Float without quotes - may cause errors
78	--sampling_temp 0.1,0.2 # Multiple temps need proper format
79	```
80
81	### Fixed Positions JSONL
82	✅ Correct:
83	```json
84	{"A": [1, 2, 3, 10, 11], "B": [5, 6]}
85	```
86
87	❌ Wrong:
88	```json
89	{"A": "1,2,3,10,11"} # String instead of list
90	{A: [1, 2, 3]} # Missing quotes on key
91	{"A": [1,2,3,]} # Trailing comma
92	```
93
94	### Chain Selection
95	✅ Correct:
96	```bash
97	--pdb_path_chains A,B # No spaces
98	```
99
100	❌ Wrong:
101	```bash
102	--pdb_path_chains A, B # Space after comma
103	--pdb_path_chains "A,B" # Quotes may cause issues
104	```
105
106	### Amino Acid Biases
107	```bash
108	# Bias toward certain AAs (positive = favor)
109	--bias_AA_jsonl '{"A": {"A": 1.5, "W": -2.0}}'
110
111	# Omit specific AAs globally
112	--omit_AAs "CM" # No cysteine or methionine
113
114	# Per-position omission
115	--omit_AA_jsonl '{"A": {"1": "C", "2": "CM"}}'
116	```
117
118	### Multi-Chain Design
119	```bash
120	# Design chains A and B together
121	--pdb_path_chains A,B
122
123	# Tie chains (same sequence)
124	--tied_positions_jsonl tied.jsonl
125	```
126
127	## Variants Comparison
128
129	\| Variant \| Use Case \| Key Difference \|
130	\|---------\|----------\|----------------\|
131	\| ProteinMPNN \| General \| Original model \|
132	\| SolubleMPNN \| Expression \| Trained on soluble proteins \|
133	\| LigandMPNN \| Small molecules \| Ligand-aware context \|
134
135	## Output format
136
137	```
138	output/
139	├── seqs/
140	│ └── backbone.fa # FASTA sequences
141	└── backbone_pdb/
142	└── backbone_0001.pdb # PDBs with designed sequence
143	```
144
145	### FASTA Header Format
146	```
147	>backbone_0001, score=1.234, global_score=1.234, seq_recovery=0.85
148	MKTAYIAKQRQISFVKSHFSRQLE...
149	```
150
151	## Common workflows
152
153	### Binder Sequence Design
154	```bash
155	python protein_mpnn_run.py \
156	--pdb_path binder_backbone.pdb \
157	--out_folder output/ \
158	--num_seq_per_target 16 \
159	--sampling_temp "0.1" \
160	--pdb_path_chains B # Design binder chain only
161	```
162
163	### Interface Redesign
164	```bash
165	# Fix core, design interface
166	python protein_mpnn_run.py \
167	--pdb_path complex.pdb \
168	--fixed_positions_jsonl core_positions.jsonl \
169	--num_seq_per_target 32
170	```
171
172	### Multi-State Design
173	```bash
174	# Design for multiple conformations
175	python protein_mpnn_run.py \
176	--pdb_path_multi state1.pdb,state2.pdb \
177	--num_seq_per_target 16
178	```
179
180	## Sample output
181
182	### Successful run
183	```
184	$ python protein_mpnn_run.py --pdb_path backbone.pdb -

People who liked this also viewed

6 picks

Type

proteinmpnn

44installs20forks

View on GitHub

How to install proteinmpnn?

adaptyvbio/protein-design-skills/proteinmpnn

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

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# ProteinMPNN Sequence Design
2
3	## Prerequisites
4
5	\| Requirement \| Minimum \| Recommended \|
6	\|-------------\|---------\|-------------\|
7	\| Python \| 3.8+ \| 3.10 \|
8	\| CUDA \| 11.0+ \| 11.7+ \|
9	\| GPU VRAM \| 8GB \| 16GB (T4) \|
10	\| RAM \| 8GB \| 16GB \|
11
12	## How to run
13
14	> First time? See [Getting started](../../docs/getting-started.md) to set up Modal and biomodals.
15
16	### Option 1: Local installation (recommended)
17	```bash
18	git clone https://github.com/dauparas/ProteinMPNN.git
19	cd ProteinMPNN
20
21	python protein_mpnn_run.py \
22	--pdb_path backbone.pdb \
23	--out_folder output/ \
24	--num_seq_per_target 16 \
25	--sampling_temp "0.1"
26	```
27
28	GPU: T4 (16GB) sufficient \| Time: ~50-100 sequences/minute
29
30	### Option 2: Modal (via LigandMPNN wrapper)
31	```bash
32	cd biomodals
33	# modal_ligandmpnn.py takes --input-pdb and forwards run.py args via --params-str
34	modal run modal_ligandmpnn.py \
35	--input-pdb backbone.pdb \
36	--params-str "--model_type protein_mpnn --number_of_batches 16 --temperature 0.1"
37	```
38
39	GPU (Modal): A10G default \| Timeout: 900s default
40
41	Note: LigandMPNN includes ProteinMPNN functionality (select with `--model_type protein_mpnn`).
42
43	## Config Schema
44
45	### Core Parameters
46
47	\| Parameter \| Default \| Range \| Description \|
48	\|-----------\|---------\|-------\|-------------\|
49	\| `--pdb_path` \| required \| path \| Single PDB input \|
50	\| `--pdb_path_chains` \| all \| A,B \| Chains to design (comma-sep) \|
51	\| `--out_folder` \| required \| path \| Output directory \|
52	\| `--num_seq_per_target` \| 1 \| 1-1000 \| Sequences per structure \|
53	\| `--sampling_temp` \| "0.1" \| "0.0001-1.0" \| Temperature (string!) \|
54	\| `--seed` \| 0 \| int \| Random seed \|
55	\| `--batch_size` \| 1 \| 1-32 \| Batch size \|
56
57	### Temperature Guide
58	```
59	0.1 -> Low diversity, high recovery (production)
60	0.2 -> Moderate diversity (default)
61	0.3 -> Higher diversity (exploration)
62	0.5+ -> Very diverse, lower quality
63	```
64
65	IMPORTANT: Temperature must be passed as a string, not float.
66
67	## Common mistakes
68
69	### Temperature Parameter
70	✅ Correct:
71	```bash
72	--sampling_temp "0.1" # String with quotes
73	```
74
75	❌ Wrong:
76	```bash
77	--sampling_temp 0.1 # Float without quotes - may cause errors
78	--sampling_temp 0.1,0.2 # Multiple temps need proper format
79	```
80
81	### Fixed Positions JSONL
82	✅ Correct:
83	```json
84	{"A": [1, 2, 3, 10, 11], "B": [5, 6]}
85	```
86
87	❌ Wrong:
88	```json
89	{"A": "1,2,3,10,11"} # String instead of list
90	{A: [1, 2, 3]} # Missing quotes on key
91	{"A": [1,2,3,]} # Trailing comma
92	```
93
94	### Chain Selection
95	✅ Correct:
96	```bash
97	--pdb_path_chains A,B # No spaces
98	```
99
100	❌ Wrong:
101	```bash
102	--pdb_path_chains A, B # Space after comma
103	--pdb_path_chains "A,B" # Quotes may cause issues
104	```
105
106	### Amino Acid Biases
107	```bash
108	# Bias toward certain AAs (positive = favor)
109	--bias_AA_jsonl '{"A": {"A": 1.5, "W": -2.0}}'
110
111	# Omit specific AAs globally
112	--omit_AAs "CM" # No cysteine or methionine
113
114	# Per-position omission
115	--omit_AA_jsonl '{"A": {"1": "C", "2": "CM"}}'
116	```
117
118	### Multi-Chain Design
119	```bash
120	# Design chains A and B together
121	--pdb_path_chains A,B
122
123	# Tie chains (same sequence)
124	--tied_positions_jsonl tied.jsonl
125	```
126
127	## Variants Comparison
128
129	\| Variant \| Use Case \| Key Difference \|
130	\|---------\|----------\|----------------\|
131	\| ProteinMPNN \| General \| Original model \|
132	\| SolubleMPNN \| Expression \| Trained on soluble proteins \|
133	\| LigandMPNN \| Small molecules \| Ligand-aware context \|
134
135	## Output format
136
137	```
138	output/
139	├── seqs/
140	│ └── backbone.fa # FASTA sequences
141	└── backbone_pdb/
142	└── backbone_0001.pdb # PDBs with designed sequence
143	```
144
145	### FASTA Header Format
146	```
147	>backbone_0001, score=1.234, global_score=1.234, seq_recovery=0.85
148	MKTAYIAKQRQISFVKSHFSRQLE...
149	```
150
151	## Common workflows
152
153	### Binder Sequence Design
154	```bash
155	python protein_mpnn_run.py \
156	--pdb_path binder_backbone.pdb \
157	--out_folder output/ \
158	--num_seq_per_target 16 \
159	--sampling_temp "0.1" \
160	--pdb_path_chains B # Design binder chain only
161	```
162
163	### Interface Redesign
164	```bash
165	# Fix core, design interface
166	python protein_mpnn_run.py \
167	--pdb_path complex.pdb \
168	--fixed_positions_jsonl core_positions.jsonl \
169	--num_seq_per_target 32
170	```
171
172	### Multi-State Design
173	```bash
174	# Design for multiple conformations
175	python protein_mpnn_run.py \
176	--pdb_path_multi state1.pdb,state2.pdb \
177	--num_seq_per_target 16
178	```
179
180	## Sample output
181
182	### Successful run
183	```
184	$ python protein_mpnn_run.py --pdb_path backbone.pdb -

People who liked this also viewed

6 picks

Type