Current highlight · 2026

MultidisciplinaryIntelligence

An evidence-linked view of molecular mechanisms, cellular state, computational models and human decision-making.

Explore the connected portfolio

Teaching and stewardship

How scientific reasoning becomes a shared practice.

Explore a teaching approach, original sample lessons and the sourced record of mentoring and scientific communication.

Explore the teaching dossier
Research direction

Current programme

Perturbation Intelligence

A multimodal Bio × AI framework for connecting perturbations, cellular state, evidence synthesis and bounded scientific-agent proposals.

  • 01Simple baseline firstComplexity is earned through validated gains.
  • 02Evidence before synthesisSource, provenance and uncertainty travel with every claim.
  • 03Human approval is mandatoryAgents retrieve and propose; they do not autonomously control research or infrastructure.

Scientific-agent architecture

Bounded coordination across evidence and tools.

  1. RAGRetrieve supported evidence

    Citations, provenance, abstention and evaluation remain visible.

  2. MCPExpose bounded capabilities

    Registered tools and resources declare inputs, outputs and permissions.

  3. A2ADelegate specialist proposals

    Explicit messages preserve task, evidence, limits and expected output.

  4. HUMANApprove the consequential step

    The final scientific decision remains inspectable, revisable and human-owned.

Evidence architecture

Evidence remains traceable across scales.

Follow the complete evidence sequence