molclaw-proteinmpnn-tool

Design and score protein sequences from PDB structures using ProteinMPNN workflows.

28|2|Updated Mar 31, 2026

Search Tags:#pdb #protein-design #sequence-scoring #structure-prediction #esmfold #proteinmpnn #protein-engineering

Ready to run

npx skills add https://github.com/InternScience/MolClaw --skill molclaw-proteinmpnn-tool

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: molclaw-proteinmpnn-tool
Source: https://github.com/InternScience/MolClaw/tree/main/skills/L1_tools/molclaw-proteinmpnn-tool
Command: npx skills add https://github.com/InternScience/MolClaw --skill molclaw-proteinmpnn-tool

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Enables automated design and scoring of protein sequences directly from PDB structures to accelerate structure-guided sequence engineering workflows and reduce manual pipeline orchestration.

Core Features & Use Cases

Structure-guided sequence design: Generate multiple candidate sequences per target using ProteinMPNN model variants with options for CA-only or soluble weights.
Scoring and validation integration: Run scoring-only mode against provided FASTA inputs and incorporate downstream structure prediction (e.g., ESMFold) for self-consistency checks.
Constraint-aware design: Support chain-specific redesign, fixed/designable residue positions, homooligomer tied positions, amino-acid omission/bias controls, and deterministic seeding for reproducibility.
Use Case: Batch-design 8 sequences for each chain in a screening set of PDB files, verify produced counts, and fetch predicted structures for validation.

Quick Start

Design eight sequences for chain A from the uploaded PDB file /path/to/input.pdb using model v_48_020 with sampling temperature 0.1 and dry_run set to False.

Frequently Asked Questions about molclaw-proteinmpnn-tool

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I design protein sequences from a PDB structure using ProteinMPNN?▼

Protein sequence design from PDB structures using ProteinMPNN involves uploading your input PDB file, selecting a model variant like v_48_020, specifying the target chain, and setting the sampling temperature to generate multiple candidate sequences.

Can I score existing protein sequences against a PDB structure without generating new designs?▼

Yes, you can run scoring-only mode against provided FASTA inputs to evaluate protein sequences against a PDB structure, allowing you to validate and score existing sequences without performing de novo sequence generation.

Does this protein design workflow support fixed residue positions and chain-specific constraints?▼

Yes, constraint-aware design supports chain-specific redesign, fixed and designable residue positions, homooligomer tied positions, amino-acid omission biases, and deterministic seeding for reproducible structure-guided sequence optimization.

What is the best way to validate ProteinMPNN designed sequences for structural consistency?▼

The best way to validate designed sequences for structural consistency is using ESMFold integration, which predicts structures from generated sequences to perform self-consistency checks against the original PDB backbone.

Can I batch process multiple PDB files for structure-guided sequence optimization?▼

Yes, batch processing for structure-guided sequence optimization supports designing multiple sequences for each chain across a screening set of PDB files, with options for CA-only and soluble model weights.