Description
Numerion Labs is a drug discovery company built around a machine learning platform for finding novel, drug-like small molecules. The company was previously known as Atomwise; the old atomwise.com domain now redirects to numerionlabs.ai.
Read More
Numerion describes its approach as exploring chemical space roughly 10,000 times larger than what competing methods reach, using three pieces of technology. COSMOS, its universal chemistry foundation model, predicts drug function directly from chemical structure and is used to consistently surface diverse, structurally novel binder candidates. APEX, a hyper-scalable enumerator, exhaustively searches that expanded chemical space for candidates. EXPO, a set of expert optimization algorithms, lets the team adapt models to a specific project quickly without needing large new datasets or retraining.
The company states these methods deliver a 10-billion-fold increase in screening efficiency and models that are roughly 100 times more efficient to optimize than earlier approaches, letting the platform search far more chemical possibilities without a matching increase in cost.
Programs built from this platform currently target immune and inflammatory diseases, with the stated goal of reaching first-in-class or best-in-class small-molecule drug candidates.
Numerion Labs is not a self-serve public tool. It is a research organization that runs its own drug discovery programs, led by a leadership team with backgrounds spanning biology, medicinal chemistry, and corporate development, alongside a board and scientific advisory board drawn from chemistry, AI, and translational science. The company publishes updates on its programs and platform through its own Insights, News, and Blog sections rather than offering an app or dashboard that outside users can sign into.
One scientist at the company describes the shift this way: everything chemists wished they could do thirty years ago, Numerion says its platform now lets them actually do, reflecting a broader claim that the company operates in underutilized, novel chemical space that older screening methods could not reach.
- CategoryScientific Discovery & Lab Tools
- DeploymentCloud
- PlatformsWeb
Numerion Labs's Core Features
Explores chemical space with a universal chemistry foundation model (COSMOS) that predicts drug function from chemical structure
Enumerates massive chemical space efficiently through a hyper-scalable enumerator (APEX)
Deploys project-specific model expertise quickly without large datasets or retraining, via expert optimization algorithms (EXPO)
Discovers structurally novel, drug-like binders rather than only variations on known chemistry
Runs internal drug discovery programs focused on immune and inflammatory diseases
Combines machine learning models with a team of chemists and engineers reviewing candidates
Aims for first-in-class and best-in-class small-molecule drug candidates
How to use Numerion Labs?
Explore chemical space: COSMOS, the universal chemistry foundation model, predicts drug function directly from chemical structure across a much larger chemical space than conventional methods.
Enumerate candidates: APEX exhaustively enumerates that chemical space to surface diverse, structurally novel binder candidates.
Optimize for the target: EXPO applies expert optimization algorithms to adapt models to a specific project without needing large new datasets.
Advance programs: Numerion's internal programs carry the strongest candidates forward, currently focused on immune and inflammatory disease targets.
Review with domain experts: A team of chemists and engineers evaluates outputs before candidates progress.
Numerion Labs's Use Cases
- Small-molecule discovery
- Immune disease programs
- Model optimization for new targets
