Tasks that previously often took weeks to complete manually can now be carried out within a few hours. At the same time, depending on the application, the system requires little or no additional training data. In a test, the detection of protein complexes achieved an F1 score of 91 per cent. The software is open-source and freely available via GitHub. It has also already been used for all tomograms in the Chan Zuckerberg Imaging Institute’s public data portal. Scalability is particularly important for biomedical research. Larger image datasets can be analysed more quickly, and results can be compared more easily across projects. This can accelerate structural biology, cell biology and drug discovery, as molecular changes associated with diseases can be investigated more systematically. MemBrain v2 was published in *Nature Methods*.
Press release from the “idw – Science Information Service" dated 8 September 2026