Numerion Labs is an AI drug hunting platform for therapeutic discovery and molecular engineering. The company was previously known as Atomwise, founded in 2015 through Y Combinator. It is based in San Francisco and led by CEO Steve Worland.
The platform centers on three named systems. COSMOS is a universal chemistry foundation model that predicts drug function from chemical structure. APEX is a hyper-scalable enumerator that virtually screens ultra-large combinatorial libraries. EXPO applies expert optimization algorithms to tailor models per project without large retraining efforts.
It targets the cost and attrition problems of traditional drug discovery. Numerion Labs states it has worked on over 600 disease targets with more than 250 pharmaceutical, agrochemical, and biotech partners. The company has raised over 174 million dollars from venture investors, according to Crunchbase.
A key differentiator is combining generative and predictive chemistry into one pipeline management environment. In October 2025, Numerion Labs published research with NVIDIA on arXiv describing APEX, which the companies say can evaluate 10 billion virtual compounds in under 30 seconds.
Pricing
Numerion Labs does not publish pricing plans, free trials, or self-serve tiers. Access is arranged through partnership and business development inquiries rather than a subscription model. No usage-based pricing or enterprise rate cards are listed publicly. Prospective partners should contact the company directly for commercial terms.
* Disclaimer: Please note that pricing information may not be up to date. For the most accurate and current pricing details, refer to the official website.
Key Features
-
✓
COSMOS foundation model predicts drug function from chemical structure
-
✓
APEX enumerator screens billions of virtual compounds in seconds
-
✓
EXPO algorithms optimize project-specific models without heavy retraining
-
✓
ADMET property prediction estimates absorption, metabolism, and toxicity
-
✓
Binding affinity modeling predicts small molecule and target interactions
-
✓
Pipeline program tracking manages progress across partner collaborations
Use Cases
De Novo Lead Generation
Medicinal chemistry teams use COSMOS to design novel small molecule candidates for specific targets. The model explores chemical space beyond existing physical compound libraries. This helps researchers surface new intellectual property early.
Ultra-Large Virtual Library Screening
Discovery teams use APEX to evaluate combinatorial synthesis libraries containing billions of compounds. The system reportedly completes this evaluation in seconds rather than months. This surfaces diverse starting points that brute-force docking would miss.
Multiparameter Lead Optimization
Research teams input early hit series to refine efficacy and pharmacokinetics together. EXPO adjusts chemical structures to satisfy competing parameters like solubility and potency. This reduces the number of physical synthesis cycles required.
In Silico ADMET Screening
Computational biologists run candidates through predictive ADMET filters before synthesis. The platform flags potential metabolic liabilities or off-target hazards early. This saves laboratory resources by eliminating non-viable molecules sooner.
Target Interaction Assessment
Biotech teams model target binding pockets to assess druggability. The platform predicts binding conformations and affinity ranges across protein structures. This lets teams focus on targets with higher computational feasibility.
Collaborative Pipeline Development
R&D leads track candidate progress across internal or partner discovery programs. The platform organizes computational data alongside target product profiles. Numerion Labs states it supports over 250 partner organizations this way.
Strengths & Weaknesses
Strengths
Pairs a proprietary foundation model, COSMOS, with hyper-scale virtual screening via APEX.
Backed by substantial venture funding, reportedly over 174 million dollars.
Established track record spanning over 600 disease targets and 250 partners.
Publishes peer-reviewed research validating its methods, including a 2025 NVIDIA collaboration.
Offers an end-to-end workflow from hit discovery through pipeline program management.
Weaknesses
Pricing and commercial licensing terms are not publicly disclosed.
No self-serve trial or public demo access exists for individual researchers.
Effective use requires specialized medicinal chemistry and computational biology expertise.
In silico predictions still require experimental wet-lab validation before advancement.
Who Is This For?
Biopharmaceutical Companies: established R&D divisions looking to add generative AI to hit-to-lead programs.
Biotech Startups: lean teams building therapeutic pipelines without large physical screening infrastructure.
Medicinal Chemists: researchers needing AI support for structural modification and synthetic feasibility checks.
Computational Biologists: scientists running virtual screening, structure-based design, and predictive ADMET modeling.
Frequently Asked Questions
Is Numerion Labs the same company as Atomwise?
Yes. Numerion Labs is the current name for the company previously known as Atomwise, founded in 2015.
Is there a free trial or self-serve plan?
No public free trial or self-serve signup is available. Access is arranged through partnership inquiries.
What is the pricing structure?
Pricing is not publicly disclosed. Commercial terms are negotiated directly with the company.
Does the platform replace wet-lab testing?
No. It generates and prioritizes computational predictions, but candidates still need experimental laboratory validation.
What are COSMOS, APEX, and EXPO?
These are the platform’s named components: a chemistry foundation model, a large-scale library screening engine, and project-specific optimization algorithms.
How large is the chemical space APEX can screen?
Numerion Labs and NVIDIA reported screening 10 billion virtual compounds in under 30 seconds in October 2025 research.
Who typically uses this platform?
Biopharma companies, biotech startups, medicinal chemists, and computational biologists working on early-stage drug discovery.
Does Numerion Labs have outside funding?
Yes. Crunchbase reports the company has raised over 174 million dollars from venture investors.
Can individuals sign up on the website directly?
No. Business inquiries and partnerships are directed through the company’s contact channels rather than direct signup.
What file formats does the platform likely support?
Specific integrations are not publicly listed, though standard chemical formats like SMILES and SDF are likely supported.
No third-party software integrations are explicitly named in public materials. The platform likely supports standard chemical computation formats such as PDB, SMILES, and SDF, but specific vendor integrations are not confirmed.