Nanobody CDR Design

Design nanobody CDR regions using BoltzGen with optimized cysteine filtering for single-domain antibodies (VHH).

Table of contents

1. Overview
2. Required MCPs
3. Configuration
4. Pipeline Steps
   1. Step 0: Setup
   2. Step 1: Prepare Configuration
   3. Step 2: Validate Configuration
   4. Step 3: Submit Design Job
   5. Step 4: Monitor Progress
   6. Step 5: Get Results
   7. Step 6: Analyze Designs
5. Output Structure
6. Available Protocols
7. Run the Workflow

---

Overview

This workflow designs nanobody complementarity-determining regions (CDRs) targeting a protein structure using BoltzGen:

1. Configuration — Create a BoltzGen YAML config specifying target structure and optional scaffolds
2. Design Generation — Run the nanobody-anything protocol for CDR loop design
3. Inverse Folding — Optimize sequences for designed backbones
4. Filtering — Cysteine filtering and quality assessment

Required MCPs

pskill install nanobody_design

This installs: boltzgen_mcp

Configuration

TARGET_CIF: "examples/data/target.cif"     # Target protein (CIF or PDB)
TARGET_CHAIN: "A"                           # Target chain ID
RESULTS_DIR: "results/nanobody_design"      # Output directory
NUM_DESIGNS: 10                             # Number of designs
BUDGET: 2                                   # Computational budget (higher = more diverse)

Pipeline Steps

Step 0: Setup

Create results directory structure.

Step 1: Prepare Configuration

Create a BoltzGen YAML configuration file specifying the target structure, chain, and optional nanobody scaffolds.

entities:
  - file:
      path: /path/to/target.cif
      include:
        - chain:
            id: A

Step 2: Validate Configuration

Verify the config file, check that target files exist and YAML structure is correct.

Step 3: Submit Design Job

Submit the job using boltzgen_submit with the nanobody-anything protocol. The protocol is specialized for:

• Single-domain antibodies (VHH)
• CDR loop design
• Cysteine filtering

Step 4: Monitor Progress

Check job status using boltzgen_check_status or boltzgen_job_status.

Step 5: Get Results

Retrieve completed designs — BoltzGen outputs CIF files in subdirectories:

• intermediate_designs/ — Backbone structures
• intermediate_designs_inverse_folded/ — Sequence-designed structures

Step 6: Analyze Designs

Summarize results, identify top candidates, and recommend designs for experimental validation.

Output Structure

RESULTS_DIR/
├── config.yaml                    # BoltzGen configuration
├── designs/
│   ├── config.cif                 # Initial config structure
│   ├── intermediate_designs/      # Backbone designs (CIF)
│   ├── intermediate_designs_inverse_folded/  # Sequence-designed
│   ├── boltzgen_run.log           # Execution log
│   └── job_info.json              # Job metadata
└── logs/

Available Protocols

Protocol	Use Case
nanobody-anything	Nanobody CDR design (default for this workflow)
protein-anything	General protein binder design
peptide-anything	Peptide binder design
antibody-anything	Full antibody design

Run the Workflow

pskill install nanobody_design
claude
> /nanobody-design