adaptyvbio/protenix
skills/protenix/SKILL.md
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, use protein-qc. For ipSAE ranking, use ipsae.
$ install --global
skillsauthnpx skillsauth add adaptyvbio/protein-design-skills protenixInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
Security Scan Results
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
3
Scanners Passed
9
Scanners in report
Clean
TrivyContainer and dependency vulnerability scanner
95%
Clean
SemgrepStatic code analysis for vulnerabilities
95%
Clean
mcp-scan (Snyk)Model Context Protocol security validation
95%
Skipped
Snyk (dep)Open source security scanning
50%
Skipped
Socket.devSupply chain security analysis
50%
Skipped
VirusTotalMulti-engine malware detection
50%
Skipped
CrowdStrikeAdvanced threat intelligence
50%
Skipped
OSV-ScannerOpen Source Vulnerability database check
50%
Skipped
OWASP Dep-Check
50%
Last scanned: Jun 12, 2026, 3:54 AM7.3s1 file scanned
SKILL.md
- name:
- protenix
- description:
- >
- skill when:
- (1) Predicting complex structures with an AF3-class model,
- license:
- MIT
- category:
- design-tools
- tags:
- [structure-prediction, validation, alphafold3, open-source]
- biomodals_script:
- modal_protenix.py
Protenix Structure Prediction
Protenix is ByteDance's open PyTorch
reproduction of AlphaFold3 (Apache 2.0). It is an AF3-class complex predictor, useful
next to boltz and chai for cross-checking designed complexes. Runnable through
biomodals.
Use Protenix-v2 for antibody-antigen complexes. The v2 model (464M params, April
2026) adds 9 to 13 percentage points of antibody-antigen accuracy over v1 at the
DockQ > 0.23 threshold and is more sample-efficient (v2 at 5 seeds exceeds v1 at 1000).
Select it with --model-name protenix-v2. For general complexes, the v1 base model is
fine.
Prerequisites
| Requirement | Value |
|-------------|-------|
| Runner | Modal (biomodals) |
| GPU | L40S (default; GPU env var) |
| Setup | See Getting started |
How to run
git clone https://github.com/hgbrian/biomodals && cd biomodals
printf '>protein|A\nMAWTPLLLLLLSHCTGSLSQ...\n' > target.faa
uv run --with modal modal run modal_protenix.py \
--input-faa target.faa \
--seeds 42 \
--no-use-msa
Key parameters
| Parameter | Default | Description |
|-----------|---------|-------------|
| --input-faa | one required | FASTA input (or --input-json) |
| --seeds | 42 | Comma-separated seeds |
| --use-msa / --no-use-msa | MSA on | Pass --no-use-msa for single-sequence |
| --model-name | v1 base | Set protenix-v2 for antibody-antigen complexes |
| --use-mini | off | Switch to the smaller protenix_mini model |
| --out-dir | ./out/protenix | Output directory |
When to use Protenix vs Boltz vs Chai
| Need | Tool | |------|------| | Affinity head (small molecules) | boltz (Boltz-2) | | Fastest, ligand support | chai | | Open AF3 reproduction | protenix (v1 base) | | Antibody-antigen complexes | protenix-v2 |
Ranking a shortlist across more than one predictor is more reliable than trusting a single model.
Troubleshooting
| Issue | Cause | Fix |
|-------|-------|-----|
| Missing input error | No --input-faa/--input-json | Provide one |
| Slow run | MSA enabled | Add --no-use-msa |
| OOM | Large complex | Use --use-mini or a larger GPU |
Next: Rank with ipsae, filter with protein-qc.
Related Skills
adaptyvbio/mosaic
devops
VerifiedTrustedCommunity
Multi-objective, gradient-based protein binder design with Mosaic. Use this skill when: (1) Composing several structure or sequence models into one design objective, (2) Optimizing binders against a custom loss rather than a fixed pipeline, (3) Wanting gradient descent over sequence space in the style of ColabDesign, RSO, or BindCraft but with interchangeable predictors, (4) Letting the optimizer choose the epitope instead of fixing hotspots. For an end-to-end binder pipeline with default filters, use bindcraft. For all-atom diffusion design, use boltzgen. For backbone-only generation, use rfdiffusion.
137SKILL.mdUpdated Jun 12, 2026
adaptyvbio/germinal
development
VerifiedTrustedCommunity
De novo antibody and nanobody (VHH) design with Germinal. Use this skill when: (1) Designing epitope-targeted nanobodies or scFvs, (2) Needing CDR design on a fixed framework, (3) Working on antibody-format binders rather than miniproteins. For miniprotein binders, use binder-design (BoltzGen, BindCraft, RFdiffusion, Mosaic). For structure validation, use boltz or chai.
137SKILL.mdUpdated Jun 12, 2026
adaptyvbio/uniprot
testing
VerifiedTrustedCommunity
Access UniProt for protein sequence and annotation retrieval. Use this skill when: (1) Looking up protein sequences by accession, (2) Finding functional annotations, (3) Getting domain boundaries, (4) Finding homologs and variants, (5) Cross-referencing to PDB structures. For structure retrieval, use pdb. For sequence design, use proteinmpnn.
137SKILL.mdUpdated Mar 27, 2026
adaptyvbio/solublempnn
development
VerifiedTrustedCommunity
Solubility-optimized protein sequence design using SolubleMPNN. Use this skill when: (1) Designing for E. coli expression, (2) Optimizing solubility of designed proteins, (3) Reducing aggregation propensity, (4) Need high-yield expression, (5) Avoiding inclusion body formation. For standard design, use proteinmpnn. For ligand-aware design, use ligandmpnn.
137SKILL.mdUpdated Mar 27, 2026