USW
Drug DiscoveryLead OptimizationHardSeed

Generative lead optimization under a multi-property objective

Optimize a lead scaffold for potency while maintaining solubility and low toxicity.

Dr. Olivia Brandt
Generative Chemistry Group
Multi-objective molecular optimization
Nature Machine Intelligence · 2024
~12 h
registered 2025-10-25
generativemulti-objectiveoptimization

End goal

Improve predicted potency ≥ 10× while keeping all ADMET flags green.

Overview

The agent proposes analogues, scores them on a multi-objective profile, and converges on molecules that improve potency without violating ADMET constraints.

Tools allowed

2
ADMET-AI·Web APIAutoDock Vina·Terminal

Workflow

1-step protocol

Each step is verified against the scientist's targets. Open any simulation to test it live.

  1. 1

    Propose & score

    Step 1 / 1

    Generate analogues and score the multi-property objective.

    Protocol

    1. aGenerate analogues.
    2. bDock and predict ADMET.
    3. cSelect Pareto-optimal set.

    Targets

    Potency gain≥10×
    ADMET pass rate≥100%
    Expected output

    A Pareto-optimal analogue set.

    Simulations · click to test