A team led by Professor Kwang-Hyun Cho of the Department of Bio and Brain Engineering has developed a generative AI system capable of predicting how cells respond to drugs or genetic modifications, and even identifying the most effective interventions to guide cells toward healthier states.  

Published in Cell Systems, the research explains how the technology models the intricate interactions between cells and drugs using a modular “Lego-block” approach, breaking them down and recombining them to forecast how a cell might behave under previously untested conditions.  

At the heart of the research lies the concept of a latent space – a mathematical map often used in image-generation AI to capture the essential features of an object or cell. The team successfully separated the representations of cell states and drug effects within this space, enabling them to recombine these elements and predict the outcomes of untested cell-drug pairs. They then extended the method to estimate how a cell’s state would change following specific genetic perturbations.  

To validate their model, the researchers applied it to colorectal cancer cells, identifying molecular targets capable of reverting malignant cells toward a normal-like state. Laboratory experiments confirmed these predictions, proving that the AI could guide cells in a desired biological direction.  

Professor Cho explained, “Inspired by image-generation AI, we applied the concept of a direction vector – a mathematical tool that enables us to transform cells in a desired direction. This technology allows quantitative analysis of how specific drugs or genes affect cells and can even predict previously unknown reactions, making it a highly generalisable AI framework.”  

This generative AI framework offers a powerful platform for designing methods to induce targeted cell-state changes, opening new possibilities for precision drug development, intelligent cancer treatments, and regenerative medicine capable of restoring damaged tissues to health.