Molecular Design
Designing bettermolecules faster
Generative models propose molecules shaped by the biology of each target, then physics-based simulation and our laboratory decide which ones deserve to exist.
- Structure-aware generationDesigns conditioned on binding-pocket geometry and interaction patterns.
- Multi-parameter optimisationPotency, selectivity, ADMET and synthesisability optimised together.
- Synthesis-aware scoringRetrosynthesis models keep designs within reach of the chemistry team.
- Molecules generated and scored in silicoPlaceholder
- 3.2M
- Assay hit rate (48 of 412 compounds)Placeholder
- 11.7%
Design cycle
Five stages, one loop
- 01
Generate
Thousands of candidate structures per target from diffusion and graph-based generative models.
- 02
Score
Predicted affinity, selectivity panels and ADMET properties rank every design.
- 03
Simulate
Docking, free-energy perturbation and molecular dynamics test the top designs physically.
- 04
Make & test
The shortlist is synthesised and assayed; results return within the same design cycle.
- 05
Learn
Active learning selects the next designs that will teach the model most.
Beyond small molecules
The same discipline for biologics and nucleic acids
For antibodies, models predict binding and flag developability risks — aggregation, polyspecificity and immunogenicity — before a single variant is expressed.
For siRNA and mRNA, sequence models optimise potency, stability and translation while penalising predicted off-target effects.
- Antibody affinity and developability
- siRNA potency and seed off-target modelling
- mRNA codon and UTR optimisation
Have a hard target?
Allosteric pockets, selectivity problems and undruggable reputations are where design matters most.