DesignGPU8 credits

ProteinMPNN — AI Protein Design

Design optimized protein sequences for any 3D backbone

ProteinMPNN uses a message-passing neural network to design protein sequences that fold into a given 3D backbone structure. It achieves 52.4% sequence recovery compared to 32.9% for Rosetta, enabling the design of thermostable, functional protein variants for any target structure.

$0.08

per API call

credits per call

/v1/design/proteinmpnn

API endpoint

Features

Inverse folding — sequence from structure

Temperature sampling for sequence diversity

Fixed position support for key residues

Multi-chain complex design

LigandMPNN variant for ligand-aware design

GPU-accelerated inference

Quick Start

ProteinMPNN — Python Examplepython

import requests

API_KEY = "sk-sci-your-key-here"
url = "https://scirouter.ai/v1/design/proteinmpnn"

response = requests.post(url, json={
    "pdb": open("backbone.pdb").read(),
    "num_sequences": 8,
    "temperature": 0.1,
    "fixed_positions": [45, 46, 47, 120, 121]
}, headers={"Authorization": f"Bearer {API_KEY}"})

result = response.json()
for seq in result["sequences"]:
    print(f"Score: {seq['score']:.3f} | {seq['sequence'][:50]}...")