binder-design - SKILL.md Agent Skill

name: binder-design description: > Guidance for choosing the right protein binder design tool. Use this skill when: (1) Deciding between BoltzGen or BindCraft, (2) Planning a binder design campaign, (3) Understanding trade-offs between different approaches, (4) Selecting tools for specific target types. license: MIT category: orchestration tags: [guidance, tool-selection, workflow] source: https://github.com/adaptyvbio/protein-design-skills

Binder Design Tool Selection

Decision tree

De novo binder design?
│
├─ Standard target → BoltzGen ✓ (recommended)
│   All-atom output, single-step: backbone + sequence + side chains
│
├─ Need diversity/exploration → BoltzGen (multiple runs)
│   or BindCraft with varied hotspots
│
├─ Integrated end-to-end → BindCraft
│   Built-in AF2 validation loop
│
├─ Ligand binding → BoltzGen ✓
│   All-atom diffusion handles ligand context natively
│
└─ Antibody/Nanobody → IgGM or mBER

Tool Comparison

Tool	Strengths	Weaknesses	Best For
BoltzGen	All-atom, single-step, ligand-aware	Higher GPU requirement (L40S)	Standard (recommended)
BindCraft	End-to-end, built-in validation	Less diverse outputs	Production campaigns
IgGM	Antibody/nanobody CDR design	Specialized format	Ab/VHH design
mBER	VHH nanobody, mask-based design	VHH-specific	VHH optimization

Recommended Pipeline: BoltzGen → Chai → QC

Target → BoltzGen → Chai → QC filter
 (pdb)  (all-atom)   (val)   (rank)

1. Target preparation

Trim to binding region + 10Å buffer
Remove waters and ligands
Renumber chains if needed

2. Hotspot selection

Choose 3-6 exposed residues
Prefer charged/aromatic residues
Cluster spatially (within 10-15Å)

3. Design with BoltzGen

# binder.yaml
entities:
  - protein:
      id: B
      sequence: 70..100
  - file:
      path: target.cif
      include:
        - chain:
            id: A
      binding_types:
        - chain:
            id: A
            binding: 45,67,89

modal run modal_boltzgen.py \
  --input-yaml binder.yaml \
  --protocol protein-anything \
  --num-designs 50

4. Alternative: BindCraft pipeline

For end-to-end design with integrated validation:

modal run modal_bindcraft.py \
  --input-pdb target.pdb \
  --hotspots "A45,A67,A89" \
  --number-of-final-designs 50

5. Validation + filtering

modal run modal_chai1.py --input-faa sequences.fasta --out-dir predictions/

Filter: pLDDT > 0.80, ipTM > 0.50, PAE_interface < 10, scRMSD < 2.0Å

Campaign Scale Guide

Stage	Count
BoltzGen designs	50-200
After Chai validation	all
After QC filtering	50-100
Experimental testing	10-50

Common Mistakes

Using buried hotspots instead of surface-exposed ones
Too many hotspots (over-constraining)
Not generating enough diversity
Including full protein instead of binding region