TLDR: A KAIST research team, led by Professor Kwang-Hyun Cho, has developed a novel generative AI technology that can predict cell responses to drugs and gene regulations. This AI models cell-drug interactions like ‘Lego blocks’ within a ‘latent space,’ enabling the identification of new drug and gene targets to guide cells toward desired states, with potential applications in cancer treatment and regenerative medicine.
SEOUL, South Korea – Researchers at the Korea Advanced Institute of Science and Technology (KAIST) have announced a significant breakthrough in artificial intelligence, developing a generative AI technology capable of predicting how cells will respond to various drugs and gene regulations. This innovative approach, led by Professor Kwang-Hyun Cho of the Department of Bio and Brain Engineering, promises to accelerate drug discovery and advance treatments for complex diseases like cancer.
The core of this new AI technology lies in its ability to mathematically model cells and drug responses by breaking them down and reassembling them, much like ‘Lego blocks.’ This modular approach allows the AI to predict not only novel interactions between cells and drugs that have not been experimentally tested but also the effects of arbitrary gene regulation. The findings were published on October 15, 2025, in the international journal ‘Cell Systems.’
Controlling the state of cells in a desired direction is a fundamental challenge in life sciences, crucial for new drug development, cancer treatment, and regenerative medicine. Traditionally, identifying suitable drugs or gene targets for this purpose has been an arduous and often unsuccessful endeavor. The KAIST team’s generative AI addresses this by operating within a ‘latent space’ – an invisible mathematical map typically used by image-generating AI to organize the essential features of objects or cells.
Within this latent space, the research team successfully separated the representations of cell states and drug effects. By recombining these representations, the AI can predict the reactions of previously untested cell-drug combinations. Extending this principle further, the model can also forecast the changes that would occur when a specific gene is regulated.
To validate their technology, the researchers utilized real experimental data. A notable success was the AI’s identification of a molecular target capable of reverting colorectal cancer cells to a state closer to normal cells. This prediction was subsequently confirmed through rigorous cell experiments. This demonstration highlights the technology’s versatility, indicating it is not limited to cancer treatment but serves as a general-purpose platform for predicting untrained transitions between various cell states and drug responses.
Also Read:
- Chemists Advocate for Ban on Generative AI in Drawing Chemical Structures Due to Error Concerns
- Stanford Researchers Unveil Paper2Agent: Transforming Scientific Literature into Interactive AI Agents
Professor Cho emphasized the significance of this achievement, stating that the technology not only determines whether a drug is effective but also reveals the underlying mechanism of how the drug functions inside the cell. This deeper understanding is critical for designing more targeted and effective therapeutic strategies. The research is expected to have broad applications in drug discovery, cancer therapy, and regenerative medicine, including the restoration of damaged cells to a healthy state.
