Academics at the École Polytechnique Fédérale de Lausanne (EPFL) have created a new tool named Systema. This new innovative tool represents a giant leap toward understanding how cell and tissue homeostasis is affected by genetic perturbations. MLBio Lab director Maria Brbic, above, is helping to lead the state with a new, cutting-edge tool. This tool addresses that head-scratching task of determining which AI model performs best in predicting these effects.
Systema was purpose-built to handle the nuances of genetic perturbation. By querying the predicted post-perturbation profiles, it learns to predict the downstream effects. This painstaking process uncovers the tangible effects of genetic alterations and offers straightforward guidance. To develop Systema, the research team conducted a thorough analysis that included data from ten other experiments to understand and guide the evolution of the program.
Ramon Viñas Torné is a postdoctoral researcher in the MLBio Lab. He is the first author of the paper describing Systema, and he was adamant about the tool’s design.
“To deal with this, we created a tool called Systema. It reduces the influence of systematic biases and focuses on the unique effects of each genetic perturbation. Systema also makes it easier to understand what genetic perturbations actually do,” – Ramon Viñas Torné.
To validate their model’s predictions, the researchers compared Systema’s performance directly against much simpler statistical approaches. The findings revealed that conventional approaches frequently outperformed the more complex AI models. This leads to concerns about how accurately these complex models actually understand genetic alterations.
Brbic spoke about the promise of emerging new technologies. These technologies have the potential to shed light on how genomic alterations dictate changes in cellular and tissue behaviors. She stated,
“The observation that simple approaches perform as well as advanced AI models made us wonder: are the advanced models actually understanding what gene changes do? Are the standard metrics suitable for evaluating these models?” – Maria Brbic.
The paper detailing the development and findings related to Systema has been published and is accessible under the DOI: 10.1038/s41587-025-02777-8.
“Looking ahead, having bigger and more diverse experiments will help make these predictions better. Also, new technologies that look at cells in more detail, like their shape or location, could help us to understand how gene changes affect cells and tissues better,” – Maria Brbic.
The paper detailing the development and findings related to Systema has been published and is accessible under the DOI: 10.1038/s41587-025-02777-8.