rowan

Rowan - Cloud Quantum Chemistry Platform

Skill Overview

Rowan is a cloud-based quantum chemistry platform that provides molecular simulation, property prediction, and protein structure prediction capabilities via a Python API, enabling complex quantum chemistry workflows without a local compute cluster.

Applicable Scenarios

1. Drug Discovery and Virtual Screening

Drug discovery scientists can use Rowan for large-scale molecular docking and property prediction. The platform supports AutoDock Vina docking, pKa prediction, solubility calculation, and more, automating virtual screening of compound libraries. All computations run in the cloud, removing the need to maintain a local HPC cluster.

2. Computational Chemistry Research

When researchers need to run DFT calculations, geometry optimizations, conformer searches, or molecular dynamics simulations, Rowan provides a unified Python API and supports multiple methods (GFN-xTB, B3LYP, AIMNet2 neural network potentials, etc.). Results can be viewed through a web interface, making it suitable for everyday quantum chemistry tasks.

3. Protein Structure Prediction and Ligand Design

Structural biologists and drug designers can use Rowan’s AI protein co-folding capabilities (Chai-1, Boltz-1/2 models) to predict protein–ligand complex structures or perform protein structure prediction. The platform handles complex steps such as protein upload and docking box definition, simplifying structure prediction workflows.

Core Features

1. Molecular Property Prediction

Rowan offers comprehensive molecular property calculations, including pKa prediction, redox potentials, solubility, and ADMET-Tox properties. It supports single-molecule and batch processing, provides RDKit-native APIs (run_pka, batch_pka), and can accept RDKit molecule objects directly to seamlessly integrate with existing cheminformatics workflows.

2. Quantum Chemistry Calculations and Geometry Optimization

The platform supports multiple tiers of quantum chemistry methods: neural network potentials (AIMNet2, Egret) for fast and accurate results; semiempirical methods (GFN1-xTB, GFN2-xTB) suitable for large systems; and DFT methods (B3LYP, ωB97X, etc.) for high-accuracy calculations. Core workflows such as geometry optimization, conformer search, and single-point energy calculations can be executed via simple API calls.

3. Protein–Ligand Docking and AI Structure Prediction

Rowan integrates AutoDock Vina for protein–ligand docking and supports fetching protein structures directly from PDB IDs. The AI protein co-folding feature uses Chai-1 and Boltz models to predict protein–ligand complex structures and outputs confidence metrics such as pTM score and interface pTM, suitable for target validation and lead optimization.

Frequently Asked Questions

Is the Rowan platform paid? How do I get an API key?

Rowan uses a credits-based billing model; new users receive a certain amount of free credits upon registration. API keys can be generated at labs.rowansci.com/account/api-keys and used for authentication via the environment variable ROWAN_API_KEY or by assigning rowan.api_key directly. Using an environment variable is recommended for security.

What quantum chemistry methods does Rowan support?

Rowan supports multiple levels of methods: neural network potentials (AIMNet2, ωB97M-D3, Egret) that approach DFT accuracy but are orders of magnitude faster; semiempirical methods (GFN1-xTB, GFN2-xTB) suitable for molecules with hundreds of atoms; and DFT functionals such as B3LYP, PBE, ωB97X with various basis sets. The platform can auto-select appropriate methods based on workflow type or you can explicitly specify them in parameters.

How do I batch-process molecular calculations in Rowan?

Rowan provides RDKit-native batch APIs, e.g., batch_pka(molecules) to process multiple RDKit molecules at once. For workflow-level batching, use batch_submit_workflow() to submit multiple jobs or batch_poll_status() to poll the status of multiple workflows. Combined with folder organization features, you can manage screening projects containing hundreds of molecules.