Unlike other generative AIs that use text or graph-based molecular representations lacking any chemical insight, our generative AI is trained on chemical reactions data that allowed it to learn organic synthetic chemistry.
By combining millions of commercial building blocks following these chemical templates, and under constraints of reward optimization, we can iteratively generate promising chemical structures, converging fast to potential drug candidates.
The unique advantages of our algorithm are:
- Generation of synthesizable compounds, by design;
- Ability to generate novel and potentially patentable compounds;
- Exploration of a huge chemical space of approximately 10²⁷ molecules;
- Multi-parametric optimization following user-defined rewards;
- Compatibility with 3D scores and constraints.