Platform

AI-driven design and automation, built for discovery

Data-efficient, physics-informed AI

We integrate physics principles in our AI architectures, creating more reliable and accurate models to venture deeper in chemical space.

NeuralPlexer

Protein-ligand structure prediction3D physics-based equivariant generative diffusion

State-of-the-art prediction of protein structures and protein-ligand complexes, including conformational response to binding, uncovering new mechanisms of action, including allostery and cryptic pockets

OrbNet

AI-accelerated quantum chemistryGraph neural network architecture based on quantum features

Protein-ligand binding energetics, 1000x faster than conventional quantum chemistry methods such as density functional theory (DFT), without compromising accuracy

PropANE

Multi-parameter lead selection

Massively pre-trained graph neural network, deployed across dozens of drug properties for lead optimization, and supported by automated training, uncertainty quantification and explainability

State-of-the-art data efficiency

Magnet

Generative molecular design

Suite of generative technologies, fully aware of the chemical space efficiently accessible through our high-throughput chemistry platform

Optimally exploring chemical space, while ensuring rapid and successful execution on our platform

New molecular designs to new biological data each week

New molecular designs to new biological data each week

Refining target profiles

Parallel testing across alternative profiles based on platform-generated biological insights

Designing drug candidates

De novo protein-ligand structure and cryptic pocket prediction

High-throughput chemistry

Nanoscale synthesis of thousands of unique compounds per week

High-throughput biology

Generating biochemical, metabolic, and cellular data

Closing the loop

Automated data pipelines and model retraining, every week

Creating the right molecule for the right profile

By identifying diverse chemical matter with unique profiles, we arrive at highly differentiated development candidates selected from multiple strong lead series.