Three EPFL researchers awarded SNSF Starting Grants in life sciences

The SNSF Starting Grants, funded by the State Secretariat for Education, Research, and Innovation (SERI), provide early-career researchers at Swiss institutions the opportunity to establish their own teams and carry out groundbreaking independent projects. Each project is selected for its potential to advance knowledge and impact the scientific community.

Camille Goemans: Combating Antibiotic Resistance with Phage-Antibiotic Combinations

Camille Goemans, head of the Goemans Lab of Drug-Microbiota Interactions at EPFL’s School of Life Sciences, focuses on the critical challenge of antibiotic resistance—a growing threat to global health. Antibiotics, once revolutionary in modern medicine, are losing efficacy due to bacterial adaptation and overuse.

Goemans’ awarded project, PACMAN, investigates the potential of combining antibiotics with bacteriophages (phages)—viruses that infect and destroy bacteria—to enhance treatment efficacy and slow the emergence of resistance.

PACMAN will use innovative, systematic approaches to study interactions between antibiotics and phages, which are currently underexplored. While phage therapy is experiencing a resurgence, it has been limited to compassionate-use cases when antibiotics fail. Goemans aims to identify and understand mechanisms of synergy between phages and antibiotics, potentially unlocking new strategies for treating bacterial infections.

The project will combine high-throughput experimental testing with advanced molecular techniques to elucidate how these combinations affect bacterial survival and adaptation. By generating a global view of the effects and interactions, PACMAN will lay the groundwork for optimizing antimicrobial therapies. This research has profound implications for addressing the antibiotic resistance crisis and could lead to transformative advancements in clinical treatments.

Mackenzie Mathis: Neural latent dynamics underlying visuomotor learning

Mackenzie Mathis, the Bertarelli Foundation Chair of Integrative Neuroscience and Director of the Mathis Laboratory of Adaptive Intelligence at EPFL’s Brain Mind Institute is tackling one of neuroscience’s foundational questions: how does the brain adapt to rapidly changing environments? It has been appreciated for centuries that our brain must build models of the world to both counter sensory delays and make predictions about our actions and those around us. Yet, the underlying neural computations that enable this adaptation are unknown.

Mathis’ awarded project focuses on visually-guided motor learning using mice. She proposed developing novel virtual reality tasks for mice, new machine learning methods, and neural recordings, to study how populations of neurons drive learning. Specifically, she will develop a suite of tasks that allow for multiple learning signals to be dissociated and then build and leverage new neural latent dynamics models in order to measure how the neural spaces changes in learning to navigate in virtual reality under uncertainty.

Beyond advancing neuroscience, the project aims to inspire new approaches in artificial intelligence by applying principles of biological learning to machine learning systems. This cross-disciplinary research bridges the gap between understanding natural intelligence and creating adaptive artificial systems.

Maria Brbić: Developing AI Models for Cellular Biology

Maria Brbić, head of the Machine Learning for Biomedicine Lab at EPFL’s School of Computer and Communication Sciences, is spearheading a transformative project at the intersection of machine learning and cellular biology. Recent advances in single-cell technologies allow scientists to study the molecular heterogeneity of individual cells, offering unprecedented insights into health and disease. However, the complexity and scale of the data require novel computational tools.

Brbić’s awarded project aims to develop a multi-modal generative AI model—a “foundation model” for cellular biology—trained on millions of cells across diverse omics datasets, including genomics, proteomics, and imaging. Unlike existing models that focus on specific modalities, this model will integrate multiple data types into a unified representation. The project’s ambitious goal is to enable new discoveries, such as identifying biomarkers, understanding gene regulation, and predicting cellular drug responses.

The project will achieve this through three key objectives: developing innovative pretraining strategies for the model, enabling discoveries in previously unexplored biological contexts, and creating a framework for predicting drug responses at a cellular level. The outcome will push the boundaries of both machine learning and biomedicine, offering new tools for understanding life at its most fundamental level and paving the way for personalized medicine.

Brbić’s research represents a paradigm shift, providing not just theoretical insights but practical tools that could revolutionize how we diagnose and treat diseases. The project’s applications are vast, spanning basic biology, pharmacology, and beyond.