Atomwise
AI superplatform that discovers novel drug-like molecules across the vast universe of chemical space
Enterprise·Technical·Powered by Proprietary ML models
Key strengths
AI-driven small-molecule drug discoveryExploration of vast chemical space for novel moleculesFocus on immune and inflammatory disease programsDeep expertise in structure-based drug designProprietary ML superplatform combining multiple predictive models
Enterprise pricing
San Francisco, USA
Founded 2012
No ratings yet
Technical Overview & Integration
Atomwise's platform is built around proprietary deep learning architectures trained on massive datasets of 3D protein–ligand complexes. Key technical aspects include:
- 3D Convolutional Neural Networks: The platform uses AtomNet®, one of the first deep learning models for structure-based drug design, which processes volumetric representations of binding sites.
- Chemical Space Coverage: The virtual screening capability spans ultra-large libraries, including make-on-demand and enumerated chemical spaces containing billions of compounds.
- Input Requirements: Typically requires a protein structure (PDB file or equivalent), a defined binding pocket, and optional reference ligand data. The team handles data preparation and docking internally.
- Output: Prioritized hit lists with predicted binding scores, drug-likeness metrics (e.g., Lipinski's Rule of Five compliance), and selectivity predictions.
- Collaborative Workflow: Atomwise engages via partnerships — technical teams share target data under an NDA; Atomwise returns screening results and iterates on SAR.
- No Public API: The platform is not available as a self-serve API; access is through structured scientific collaboration agreements.