Genesis Therapeutics

Our PLatform

Discovering Novel Molecular Scaffolds for Previously Undruggable Targets

Genesis Therapeutics has created the industry's most advanced molecular AI platform by combining 3D structure-aware deep neural networks, new molecular simulation methods, and a massively scalable molecular generation engine.

Our technology

Machine Learning

Our Dynamic PotentialNet platform uniquely intuits the 3D structural dynamics of protein-ligand complexes. This gives us unprecedented predictive accuracy for potency and selectivity when drugging challenging, data-poor protein targets.
‍
We train our state-of-the-art ML models with proprietary datasets comprised of millions of potency + ADMET datapoints extensively vetted for assay quality. These massive multitask models, combined with proprietary Active Transfer Learning techniques, are uniquely suited to explore untapped areas of chemical space and identify novel candidates for previously undruggable targets.

Molecular Simulation

To unlock the most challenging protein targets, we've deeply integrated our ML models with our own cutting-edge molecular simulation platform.
‍
Drawing on decades of research in molecular dynamics and quantum mechanics, we've developed entirely new simulation methods that elucidate the flexibility, solvation energy, and binding kinetics of novel proteins. We've found that this ML/MD synthesis is crucial for understanding nuances of these high-value targets.

Molecular Generation

We developed the GEMS (Genesis Exploration of Molecular Space) platform in a close collaboration between our chemists, ML researchers, and software engineers, ensuring that our generated molecules serve our chemists' needs at every stage of the drug discovery process.
‍
During hit ID, GEMS generates billions of virtual molecules optimizing for lead-likeness, easy synthesizability, and diversity, in order to maximally explore the most promising patches of chemical space. During hit-to-lead and lead-optimization, our chemists direct GEMS to modify a starting set of hit or lead scaffolds, with a variety of settings available for larger scaffold-hopping changes or close-in lead-op scenarios. This allows us to drive towards drug candidates well-optimized for all potency + ADMET parameters with unprecedented speed and efficiency.

Publications

Our academia-leading research was the initial component of our expanding portfolio of machine learning and molecular simulation technologies

1

Our AI platform expands on the initial discoveries in PotentialNet — field-leading, peer-reviewed methods for molecular property prediction that co-founder Evan Feinberg invented in Stanford's acclaimed Pande lab.

Key components of our platform have been rigorously tested against current state-of-the-art methods in a collaboration between Stanford and a top-five pharma company. We achieved a step-change improvement predicting 20+ different ADMET properties.