MolPE 分子势能
中EN
Request a trial

Molecule design, simpler with AI

Connecting sequence, structure and function to expand molecular design.

Explore MolEX

Not a tool, but your AI research colleague

On its own, it breaks down tasks, plans experiments, calls tools and coordinates optimization.

Understand intent, execute autonomously

  • Intelligent task breakdown and planning
  • Multi-agent collaborative execution
  • Hierarchical decisions with reflective refinement

3D structure editing, in sync with the agent

  • Residue selection and candidate comparison
  • Design Copilot interprets in sync
  • Collaborative annotation by members

Empowered by knowledge and tools

  • Targeted learning from a private knowledge base
  • In-house and open-source models integrated
  • Toolchain extended through plug-ins

Build your own skills and workflows

  • Drag-and-drop workflow nodes
  • Upload and manage Skills
  • Team SOPs as executable chains

Readable output, traceable process

  • Multiple sequence alignment view
  • One-click export of data, figures and reports
  • Layered memory, a traceable process
  • AlphaFold 3structure
  • ESM-2language
  • RFdiffusiondesign
  • FoldXevaluation
  • Boltz-2structure
  • ProteinMPNNdesign
  • Rosettaevaluation
  • AntiBERTylanguage
  • Chai-1structure
  • BindCraftdesign
  • ThermoMPNNevaluation
  • ESM Clanguage
  • ColabFoldstructure
  • PepMLMdesign
  • FreeSASAevaluation
  • ProtGPT2language
  • IgFoldstructure
  • DiffDockstructure
  • GearBindevaluation
  • AbLang-2language
  • ABodyBuilder2structure
  • OpenFoldstructure
  • OpenStructureevaluation
  • ESMFoldstructure
  • HADDOCKstructure
  • ImmuneBuilderstructure
  • AlphaFold 2structure

All models in the model hub

Fab–TNFα binding review
Continuous playback

Agent session: task breakdown, sub-agents in parallel, a plan revised after reflection

Sequence, structure, function

  1. Sequence

    Generate candidates in sequence space

    Combine generative models and evolutionary learning to explore protein and peptide sequences around a functional objective, and propose candidate molecules.

  2. Structure

    Understand binding in three dimensions

    Connect amino acid sequence, conformation and function, and see how the CDR loops meet the target surface.

  3. Function

    Balance many properties at once

    A world model keeps sequence, function and developability in one view, and weighs several design objectives while it generates.

  • Activity
  • Stability
  • Developability
  • Safety

Binding is not the same as being developable.

The usual route secures binding first, then screens one property at a time; one failure sends it back to the start. Our world model captures how the properties couple, and weighs them all as it generates.

  1. Sequential optimization

    Generate binders first, then screen one property at a time. Each round drops more candidates and narrows the feasible space; a failure at the last step sends the work back to the start.

  2. Joint multi-property design

    The world model evaluates every property while it generates, and each round moves all the candidates toward the region where every one of them holds.

Nearly 100%

Multi-property optimization gain over existing SOTADualGPT-AB · CDRH3 multi-property generation tasks

Over 50%

Data cost reductionDualGPT-AB · less reliance on multi-property labels

DualGPT-AB · Nature Computational Science (2026)

From platform to delivery

Run your own research on the platform, build on our data, or hand the project to us.

  • Platform

    MolEX, your AI research colleague

    It breaks down tasks, plans experiments, calls tools and optimizes with you.

    Multi-agent collaborationKnowledge and toolsMemory and reasoning

    Explore the platform

  • Data platform

    Connected biological data

    Sequence, structure, omics and scientific literature brought together to inform design and analysis.

    1.14M+

    Molecule–target pairs

    2,527

    Targets covered
    Explore the data platform

  • Solutions

    AI CRO, delivered by project

    The dry-lab service delivers intelligently designed molecules; the end-to-end service runs from target to PCC and delivers experimentally validated candidates.

    2–4 weeks

    AI-driven dry-lab CRO serviceMode 1 · delivers intelligently designed future molecules

    6–8 weeks

    AI-driven end-to-end early research serviceMode 2 · target to PCC, delivers experimentally validated candidates
    Explore solutions

From research methods to applied discovery

All cases

Take the next step in your research

Request a trial or Book a consultation. Tell us what you are working on.

Contact us