A dynamic gut digital twin

Context and objectives

For a global health and nutrition R&I leader, we built a biological digital twin of the gut microbiome, paired with an agentic AI assistant, so that scientists can design, run and interpret in-silico experiments (diets, probiotics, small molecules, etc.) in plain language before committing to wet-lab work. The aims were to:

• compress experimentation cycles from months of wet-lab work to hours of in-silico exploration;
• empower R&I scientists to run virtual what-if scenarios on diets, probiotics and small molecules in plain language.

Outcome

• A working biological digital twin of the gut microbiome, paired with a conversational AI agent.
• Scientists run microbiome experiments, designing diet, probiotic and small-molecule scenarios.
• A foundation for new R&I strategies, with mechanistic and AI models validated on scarce datasets, ready for academic-grade publication and scale-up.

Our approach

Digital twin and data foundations

• A dynamic gut digital twin built on proprietary raw DNA, metabolic models and nutrition data
• A six-step pipeline from ingestion to simulation, including diet optimisation

Conversational interface and agentic layer

• Chat-based experiment set-up with a 3D simulation view
• An agentic layer that captures intent and orchestrates the pipelines

Scientific delivery and deployment

• Workshops to define the vision, features and roadmap of strategic use cases
• Development and testing of a proof of concept and an MVP, deployed on HPC cloud infrastructure
• Development of state-of-the-art computational biology tools

A team combining PhDs in biology and applied mathematics, working on Azure, Databricks and Claude with MCP.