boltz-protein-design

name	boltz-protein-design
description	Generate novel protein binders such as peptides, antibodies, nanobodies, or custom proteins for a target with the Boltz Compute API and return ranked designed sequences with predicted complex structures. TRIGGER when the user wants to design or invent new binders for a target. Not for screening user-supplied proteins and not for small molecules.

Workflow

Use this skill when the user wants novel binder generation.

1. Normalize the target as structure_template or no_template.
2. Normalize the binder specification as either structure_template redesign or no_template sequence-DSL design.
3. Keep rules and motif constraints at defaults unless the user explicitly asks for them.
4. Estimate cost first, show the estimate, and ask for confirmation before running the design job.
5. Let the wrapper poll, paginate designs, rank outputs, and download structures locally.

Always Do This

• Use the Python SDK wrapper only: python scripts/query.py ....
• Expect authentication through BOLTZ_API_KEY; BOLTZ_COMPUTE_API_KEY also works.
• Always run --estimate-only before submission and ask for explicit cost approval.
• Always send modifications: [] on every protein, RNA, or DNA entity when there are no modifications. The API currently rejects omitted polymer modifications.
• Treat --num-proteins 10 as the minimum valid request size. Smaller values currently fail server-side.
• Treat all residue indices as 0-based.
• Be careful with the sequence DSL for designed_protein entities. Fixed residues stay literal; numeric spans represent designed segments.
• Mention optional amino-acid and motif rules to the user, but apply them only on request.
• Expect downloads under $BOLTZ_OUTPUT_DIR unless --output-dir is explicitly provided.

Input Normalization

• --binder-sequence is the simplest no-template path for designed binders.
• --binder-structure plus --binder-chain-selection is the structure-template redesign path.
• --target-protein accepts raw sequence, FASTA, or .txt.
• --target-structure accepts a local structure path or URL.
• Use --config for advanced motif layouts, rules, target epitope definitions, bonds, or constraints.

Read This Reference When Needed

• Read references/api.md for:
  • binder specification variants
  • sequence DSL semantics
  • redesign motif semantics
  • target variants and rule fields
  • returned metrics and downloaded artifacts

Command Pattern

Estimate first:

python scripts/query.py \
  --num-proteins 50 \
  --modality peptide \
  --binder-sequence "8..15" \
  --target-protein TARGET.fasta \
  --estimate-only

Submit after confirmation:

python scripts/query.py \
  --num-proteins 50 \
  --modality peptide \
  --binder-sequence "8..15" \
  --target-protein TARGET.fasta

Outputs

• results.json
• results.csv
• per-design structures under $BOLTZ_OUTPUT_DIR/<job_id>/structures/

Report the strongest designs and tell the user where the artifacts were written under $BOLTZ_OUTPUT_DIR.