protenix

Name: protenix
Author: adaptyvbio

7installs20forks

How to install protenix?

adaptyvbio/protein-design-skills/protenix

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

Installs into the current project.

Summary

Structure prediction with Protenix, an open AlphaFold3 reproduction. Use this skill when: (1) Predicting complex structures with an AF3-class model, (2) Wanting an open alternative to AF3 alongside Boltz and Chai, (3) Validating designed binder-target complexes. For QC thresholds

Files · 1

View on GitHub

SKILL.md

1	# Protenix Structure Prediction
2
3	[Protenix](https://github.com/bytedance/Protenix) is ByteDance's open PyTorch
4	reproduction of AlphaFold3 (Apache 2.0). It is an AF3-class complex predictor, useful
5	next to `boltz` and `chai` for cross-checking designed complexes. Runnable through
6	biomodals.
7
8	Use Protenix-v2 for antibody-antigen complexes. The v2 model (464M params, April
9	2026) adds 9 to 13 percentage points of antibody-antigen accuracy over v1 at the
10	DockQ > 0.23 threshold and is more sample-efficient (v2 at 5 seeds exceeds v1 at 1000).
11	Select it with `--model-name protenix-v2`. For general complexes, the v1 base model is
12	fine.
13
14	## Prerequisites
15
16	\| Requirement \| Value \|
17	\|-------------\|-------\|
18	\| Runner \| Modal (biomodals) \|
19	\| GPU \| L40S (default; `GPU` env var) \|
20	\| Setup \| See [Getting started](../../docs/getting-started.md) \|
21
22	## How to run
23
24	```bash
25	git clone https://github.com/hgbrian/biomodals && cd biomodals
26
27	printf '>protein\|A\nMAWTPLLLLLLSHCTGSLSQ...\n' > target.faa
28
29	uv run --with modal modal run modal_protenix.py \
30	--input-faa target.faa \
31	--seeds 42 \
32	--no-use-msa
33	```
34
35	## Key parameters
36
37	\| Parameter \| Default \| Description \|
38	\|-----------\|---------\|-------------\|
39	\| `--input-faa` \| one required \| FASTA input (or `--input-json`) \|
40	\| `--seeds` \| `42` \| Comma-separated seeds \|
41	\| `--use-msa` / `--no-use-msa` \| MSA on \| Pass `--no-use-msa` for single-sequence \|
42	\| `--model-name` \| v1 base \| Set `protenix-v2` for antibody-antigen complexes \|
43	\| `--use-mini` \| off \| Switch to the smaller `protenix_mini` model \|
44	\| `--out-dir` \| `./out/protenix` \| Output directory \|
45
46	## When to use Protenix vs Boltz vs Chai
47
48	\| Need \| Tool \|
49	\|------\|------\|
50	\| Affinity head (small molecules) \| boltz (Boltz-2) \|
51	\| Fastest, ligand support \| chai \|
52	\| Open AF3 reproduction \| protenix (v1 base) \|
53	\| Antibody-antigen complexes \| protenix-v2 \|
54
55	Ranking a shortlist across more than one predictor is more reliable than trusting a
56	single model.
57
58	## Troubleshooting
59
60	\| Issue \| Cause \| Fix \|
61	\|-------\|-------\|-----\|
62	\| Missing input error \| No `--input-faa`/`--input-json` \| Provide one \|
63	\| Slow run \| MSA enabled \| Add `--no-use-msa` \|
64	\| OOM \| Large complex \| Use `--use-mini` or a larger GPU \|
65
66	---
67
68	Next: Rank with `ipsae`, filter with `protein-qc`.

People who liked this also viewed

6 picks

Type

protenix

7installs20forks

View on GitHub

How to install protenix?

adaptyvbio/protein-design-skills/protenix

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

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Protenix Structure Prediction
2
3	[Protenix](https://github.com/bytedance/Protenix) is ByteDance's open PyTorch
4	reproduction of AlphaFold3 (Apache 2.0). It is an AF3-class complex predictor, useful
5	next to `boltz` and `chai` for cross-checking designed complexes. Runnable through
6	biomodals.
7
8	Use Protenix-v2 for antibody-antigen complexes. The v2 model (464M params, April
9	2026) adds 9 to 13 percentage points of antibody-antigen accuracy over v1 at the
10	DockQ > 0.23 threshold and is more sample-efficient (v2 at 5 seeds exceeds v1 at 1000).
11	Select it with `--model-name protenix-v2`. For general complexes, the v1 base model is
12	fine.
13
14	## Prerequisites
15
16	\| Requirement \| Value \|
17	\|-------------\|-------\|
18	\| Runner \| Modal (biomodals) \|
19	\| GPU \| L40S (default; `GPU` env var) \|
20	\| Setup \| See [Getting started](../../docs/getting-started.md) \|
21
22	## How to run
23
24	```bash
25	git clone https://github.com/hgbrian/biomodals && cd biomodals
26
27	printf '>protein\|A\nMAWTPLLLLLLSHCTGSLSQ...\n' > target.faa
28
29	uv run --with modal modal run modal_protenix.py \
30	--input-faa target.faa \
31	--seeds 42 \
32	--no-use-msa
33	```
34
35	## Key parameters
36
37	\| Parameter \| Default \| Description \|
38	\|-----------\|---------\|-------------\|
39	\| `--input-faa` \| one required \| FASTA input (or `--input-json`) \|
40	\| `--seeds` \| `42` \| Comma-separated seeds \|
41	\| `--use-msa` / `--no-use-msa` \| MSA on \| Pass `--no-use-msa` for single-sequence \|
42	\| `--model-name` \| v1 base \| Set `protenix-v2` for antibody-antigen complexes \|
43	\| `--use-mini` \| off \| Switch to the smaller `protenix_mini` model \|
44	\| `--out-dir` \| `./out/protenix` \| Output directory \|
45
46	## When to use Protenix vs Boltz vs Chai
47
48	\| Need \| Tool \|
49	\|------\|------\|
50	\| Affinity head (small molecules) \| boltz (Boltz-2) \|
51	\| Fastest, ligand support \| chai \|
52	\| Open AF3 reproduction \| protenix (v1 base) \|
53	\| Antibody-antigen complexes \| protenix-v2 \|
54
55	Ranking a shortlist across more than one predictor is more reliable than trusting a
56	single model.
57
58	## Troubleshooting
59
60	\| Issue \| Cause \| Fix \|
61	\|-------\|-------\|-----\|
62	\| Missing input error \| No `--input-faa`/`--input-json` \| Provide one \|
63	\| Slow run \| MSA enabled \| Add `--no-use-msa` \|
64	\| OOM \| Large complex \| Use `--use-mini` or a larger GPU \|
65
66	---
67
68	Next: Rank with `ipsae`, filter with `protein-qc`.

People who liked this also viewed

6 picks

Type