Skip to content
Scientist reviewing a protein structure and molecular design dashboards on large screens

AI & Computational Science

AI-drivendrug discovery

We fuse machine learning with physics-based simulation and real experimental data to identify and optimise therapeutics faster — and with a clearer view of the risks.

  • Learned modelsGraph, transformer and diffusion models for molecules, proteins and cells.
  • Physics-based simulationFree-energy methods and molecular dynamics where accuracy matters most.
  • Calibrated uncertaintyEvery prediction carries a confidence estimate that guides the next experiment.
Our technology

Methods

The toolbox

  • Knowledge graphs

    Relationships between genes, pathways, diseases, compounds and trials, queryable and explainable.

  • Omics models

    Representation learning on bulk and single-cell data to detect disease states and drivers.

  • Generative chemistry

    Pocket-conditioned diffusion and graph generation with multi-objective optimisation.

  • Protein and antibody models

    Structure prediction and sequence models for binding and developability.

  • Molecular simulation

    Docking, free-energy perturbation and long-timescale molecular dynamics.

  • Active learning

    Choosing the experiments that reduce model uncertainty most per assay.

Principle

Models earn trust at the bench

We benchmark models prospectively — on experiments run after the prediction was made — rather than only on retrospective test sets, which tend to flatter performance.

Negative results are kept, labelled and used. They are often the most informative data a model sees.

  • Prospective validation
  • Calibrated confidence
  • Full experiment lineage

Computational scientist? Come build with us.

We hire across machine learning, cheminformatics, structural biology and data engineering.

Careers