Atomwise
AI superplatform that explores chemical space to discover novel, drug-like molecules
Enterprise·Technical·Powered by Proprietary ML (AtomNet)
Key strengths
AI-driven small-molecule drug discoveryExploration of vast chemical space beyond traditional screeningFocus on first- and best-in-class immune/inflammatory disease programsProprietary deep learning models for molecular binding predictionIntegrated superplatform combining ML with medicinal chemistry
Enterprise pricing
San Francisco, USA
Founded 2012
No ratings yet
Technical Integration & Platform Details
Atomwise's platform is not exposed via a public API; engagement occurs through structured research collaborations. Key technical aspects include:
- AtomNet Architecture — A deep convolutional neural network that operates on 3D voxelized representations of protein-ligand complexes, enabling structure-based virtual screening at scale.
- Chemical Space Coverage — The platform can virtually screen libraries exceeding billions of enumerated compounds, far surpassing the ~10 million compounds in typical physical HTS libraries.
- Input Requirements — Partners typically provide a protein crystal structure (PDB format) or homology model, along with defined binding site coordinates. Apo or holo structures are both acceptable inputs.
- Output Deliverables — Ranked lists of candidate molecules with predicted pIC50/binding scores, ADMET property flags, and synthetic accessibility annotations; often delivered in SDF/CSV format.
- Generative Chemistry — Beyond screening, the superplatform includes generative ML modules for de novo molecule design around a given pharmacophore or scaffold.
- Collaboration Models — Engagements range from fee-for-service virtual screens to deep co-development partnerships with milestone-based agreements.