Endocytosis, a complex cellular process crucial for absorbing nutrients, signaling molecules, viruses, and drugs, is coordinated by thousands of protein molecules. Due to the complexity of the process, endocytosis is neither fully understood nor visualized with sufficient resolution. Jonas Ries, group leader at the Perutz since 2023, coordinates the interdisciplinary project ‘Dynamic nanoscale reconstruction of endocytosis with high-throughput super-resolution microscopy and machine-learning’ – in short: ‘DynRec’.
The project’s goal is to visualize structural changes during endocytosis at a mechanistic level. “To perform high-throughput measurements in tens of thousands of yeast cells, we use super-resolution microscopy and develop machine learning techniques to assemble snapshots of cellular processes into a movie”, Jonas Ries explains. From Jonas’ perspective, including deep learning models in basic research is groundbreaking: “At the moment, we cannot even start to comprehend to what extent it will impact our research. In this project, we use AI/ML to analyze large data sets faster and more precisely than what is possible for a human to do manually.”
Secondary structure motifs in messenger RNAs are known to play a regulatory role in translation and mRNA stability. However, the sequence features and functions of these motifs largely remain to be elucidated. Perutz group leader Stefan Ameres, CO-PI of the project ‘Determinants of mRNA Lifetime and Translation Efficiency’ – short ‘riboAI’ –, investigates new and relevant determinants of mRNA lifetime and translation efficiency. Together with coordinator Ivo Hofacker and CO-PI Sebastian Tschiatschek, both from the University of Vienna, the project will train deep learning models with RNA sequencing datasets to predict a mRNA’s lifetime and translation efficiency. “In our lab, we are going to test the extracted sequence features using a high throughput approach to experimentally validate the importance of the discovered sequence features”, Stefan Ameres explains.
The project is the foundation for future projects: “With deep learning, we have the unique possibility to use the enormous wealth of available data and extract information that would have otherwise escaped our attention”, Stefan says. “In the future, we not only want to learn more about the mechanisms of gene expression control, but also apply this knowledge to design synthetic mRNAs that can be used for therapeutic purposes, such as mRNA vaccines.”