Optagon Labs

Bringing drugs to technical
maturity, faster and cheaper.

An AI-native R&D company helping drug developers co-optimize multiple drug properties in a fraction of the time and cost, and de-risk the path to clinical trials.

See our approach Contact us

To be viable, a drug molecule must be optimized for many properties at the same time. Today this process entails siloed experiments run over many years, and success hinges on the combination of design parameters you select and lock in early.

Optagon redesigns this process and orchestrates an optimized sequence of experiments, leveraging digital twins that understand how changes to the molecule affect all properties and how to converge rapidly on the optimal design.

  1. The system

    The digital twin

    Envelops the drug lead: molecule, formulation, process, and the body it must survive in.

  2. Co-optimize multiple properties

    Find the parameters that satisfy every target at once.

  3. De-risk the next milestones

    Bring tomorrow's constraints into today's design.

  4. Accelerate the path to clinical trials

    Run only the experiments that resolve the most uncertainty.

Our AI-native R&D services accelerate drug development by running design–make–test loops across your internal and external labs. Every decision is interpretable, and carries its uncertainty and its provenance with it.

The parameters behind potency, ADMET and CMC are coupled, so they have to be tuned together: first by bringing them into one digital twin, then by running the few high-value experiments that calibrate it.

Worked example: an oral small molecule inhibitor. Hover a stage to see the parameters that move it and the targets it must hit.

Frontier model

One system that understands the drug on every level.

Three components, trained and deployed as one: a molecular foundation model, a state representation architecture for the system it lives in, and an optimizer that decides which experiment is worth running next.

  1. Molecular foundation model

    Maps molecular behaviour to molecular function.

    Breakthrough

    Frontier performance in very low data regimes, where the measurements that exist are few and expensive.

  2. State representation architecture

    Models the mechanisms that govern the drug in formulation, in transport, and in the body.

    Breakthrough

    A state-of-the-art representation that carries multi-level context and, through our frontier inference-time compute algorithms, understands which actions lead to which outcomes.

  3. Adapted Bayesian optimization

    Optimizes over the state representation, so every experiment sharpens the whole twin.

    Breakthrough

    A novel multi-level surrogate. It starts from strong physics priors, then needs only a few experiments to populate the state representation and converge on sharper posteriors, fast.

Engagement

From the properties you need to a system you can keep tuning.

One engagement, scoped to the stage you are in.
The twin stays with the program afterwards.

  1. 01

    Bring the properties you need to hit at once.

    Stability, toxicity, half-life, permeability, and more.

  2. 02

    Deploy a digital twin of your drug.

    It understands what moves those properties.

  3. 03

    Optagon directs the experimentation.

    Each experiment chosen to reduce cost and time.

  4. 04

    You keep a tunable system.

    If a property drifts in the future, you can fix it fast.

Let's talk about your program.

Bring the properties you need to hit, and we will scope where a digital twin resolves the most risk first.

Talk to us