New Biomodel Aims to Improve Understanding of Cell Behavior

The Chan Zuckerberg Initiative (CZI) has introduced GREmLN, an artificial intelligence model designed to help researchers better understand the networks that govern cell behavior. GREmLN, which stands for Gene Regulatory Embedding-based Large Neural model, is part of CZI’s ongoing effort to develop AI biomodels that can predict and explain cellular function from the molecular level up to entire biological systems.

GREmLN’s primary goal is to reveal how genes within a cell interact and what might cause these interactions to malfunction, leading to diseases such as cancer or neurodegeneration. According to Andrea Califano, president of the Chan Zuckerberg Biohub New York, “GREmLN is a novel approach to understanding how cells make decisions and, just as importantly, how those decisions go wrong in diseases like cancer.” Califano emphasized that the model is designed to align with biological realities, stating, “This model doesn’t try to reshape biology to fit AI, it reshapes AI to fit biology.”

Unlike typical AI models, GREmLN is centered on the “molecular logic” that defines gene interactions, likened to a conversation occurring within the cell. This focus enables scientists to identify early changes that may indicate disease onset and to pinpoint potential targets for new therapies.

The model is trained on over 11 million data points from CellxGene, a tool that supports thousands of scientists in comparing and analyzing single-cell data from tissues such as the brain, lung, kidney, and blood. GREmLN is expected to scale up to address significant biological and medical questions, including detecting early signs of cancerous changes in cells or initial damage in neurons. 

“Understanding cellular behavior means understanding the network of conversations happening inside every cell,” said Theofanis Karaletsos, senior director of AI at CZI. “GREmLN captures that complexity in a way we’ve never been able to before. It’s a step toward building systems that help us simulate and predict the behavior of cells.”