Property prediction at the boundary of physics and ML

The microstate population by pH of Xeljanz, an autoimmune-disease-targeting prescription medication

Macroscopic pK_a

Predict macroscopic pK_a values, microstate populations, isoelectric points, and logD values with Rowan's macroscopic pK_a workflow. This is made possible by Starling, a physics-informed machine learning model that runs in minutes. Read the preprint.

Quick conformer searching

Quickly explore conformational space using fast low-level methods for conformer generation and more accurate final methods for conformer ranking.

Regioselectivity and reactivity

Predict where a molecule will react with nucleophiles, electrophiles, and radicals and quantify an electrophile's ability to promote covalent reactions with Fukui index and global electrophilicity index calculations.

Blood–brain-barrier permeability

Predict the likelihood of blood–brain-barrier penetrance in silico by computing the free energy of neutralization and energy of solvation.

More property predictions

Predict bond-dissociation energies (BDE), solid organic solubility, hydrogen-bond-acceptor strength, redox potentials, and more. Read more about our platform.

Molecular simulation millions of times faster

Egret-1

Egret-1 is a family of open-source neural network potentials that match or exceed the accuracy of quantum-mechanics-based simulations while running orders-of-magnitude faster. With Egret-1, scientists can quickly get trustworthy results from computation to guide their work. Read the preprint.

AIMNet2

AIMNet2 is a generally applicable, accurate, and incredibly fast neural network potential that powers organic-focused computational chemistry simulations in Rowan.

OMol25 eSEN

OMol25's eSEN Conserving Small from Meta FAIR is a model trained on Open Molecules 2025, a high-quality dataset of unprecedented scale spanning small molecules, biomolecules, metal complexes, and electrolytes, including 83 elements, charged systems, and open-shell species.

Orb-v3

Orb-v3 from Orbital Materials can scale to simulations of 100,000 atoms while performing an energy and force evaluation in under 1 second.

Traditional physics-based methods

Run density-functional theory (DFT) and xTB methods with a unified interface, deployment environment, and database for calculation submission, management, and analysis.

Deep learning and physics for protein–ligand complex modeling

Compounds redocked into a protein with Rowan's strain-corrected docking worfklow

Strain-corrected docking with Vina

Test binding, generate bound poses, and search through chemical libraries with ligand-strain-corrected docking powered by AutoDock Vina.

The barnase–barnstar complex predicted from protein sequences by Boltz-2

Co-folding with Boltz-2 and Chai-1r

Predict the 3D structures and binding affinities of protein–ligand complexes from sequence information with state-of-the-art models Boltz-2, Chai-1r, and Boltz-1.

Secure infrastructure

We offer single-tenant and customer-managed virtual-private-cloud (VPC) deployments for enterprise accounts. Read more about our security and deployment options.

Python and RDKit APIs

Run calculations and workflows programmatically with a Python- or RDKit-native API that returns structured data, perfect for high-throughput screening or complex workflows. Read more about our API.

Web-native platform

Our no-code web-based platform makes it easy to submit, view, and analyze complex calculations. All workflows automatically generate publication-quality visuals, and Rowan makes it easy to securely share and collaborate within organizations.