A team of researchers from Helmholtz Munich, the Technical University of Munich (TUM), and the Biozentrum at the University of Basel has created a new artificial intelligence (AI) tool called MemBrain v2. This software automates the analysis of 3D images of cell membranes, greatly reducing the time needed for such tasks. The study, published in the journal Nature Methods, describes how the freely available tool can identify cell membranes, locate specific proteins within them, and analyze how these proteins are arranged. This provides researchers with detailed insights into how cells function at the molecular level. The AI requires very little or no extra training data, making it easier for scientists around the world to study cells in greater detail and on a larger scale. Cryo-electron tomography (cryo-ET) is a microscopic imaging technique that enables researchers to look at cells in three dimensions with high resolution. This is done by rapidly freezing the cells to preserve their structure. However, analyzing the membranes in these images has been difficult due to technical limitations that create gaps in the data. MemBrain v2 helps overcome this challenge by automating the analysis process. The tool has two main components: MemBrain-seg, which identifies membranes without needing any extra data, and MemBrain-pick, which requires only a small amount of training data. In a test, researchers manually labeled the positions of protein complexes on one membrane, and the AI correctly located the same complexes on other membranes with an accuracy score of 91%. In the past, analyzing 3D cell images required researchers to label data manually, a process that was both time-consuming and rarely useful for new data sets. Existing software usually handled only parts of the analysis, such as outlining membranes or finding proteins within them. MemBrain v2 combines three key steps into one tool: identifying membranes (MemBrain-seg), locating proteins in the membranes (MemBrain-pick), and measuring how these proteins are arranged (MemBrain-stats). In several tests, the AI produced results that matched manual analysis but was much faster. It is also user-friendly and can be adapted to different research questions with minimal changes. Since all parts of MemBrain v2 are open source, the membrane-detection component is already being used widely around the world. It has been applied to various data sets from the Chan Zuckerberg Imaging Institute, contributing to new discoveries in biology. For example, the tool revealed that certain photosynthesis-related proteins are spatially separated within the membrane, which challenges previous models of how these proteins are organized. In the future, the AI is expected to distinguish between different types of proteins even more accurately. The study was led by first author Lorenz Lamm and senior author Dr. Tingying Peng.