Teaching Cells to Talk: Decoding RNA Biochemistry From Condensate Images
Event details
| Date | 15.12.2025 |
| Hour | 10:00 › 11:00 |
| Speaker | Anita Ðonlić, Ph.D., Princeton University, NJ (USA) |
| Location | Online |
| Category | Conferences - Seminars |
| Event Language | English |
3-DAY BIOE MINI-SYMPOSIUM on Life Science Engineering
(DAY TWO: talk three / previous talk / next talk)
Abstract:
Biomolecular condensates serve as dynamic hubs for RNA-centered processes, including rRNA synthesis, pre-mRNA splicing, and viral RNA replication. However, linking condensate mesoscale morphology to the underlying biochemical activity of these RNA pathways has remained a fundamental challenge. Being able to infer RNA pathway disruptions directly from condensate imaging would enable quantitative decoding of condensate function, illuminate mechanisms of drug action, and transform microscopy data into predictive, functional readouts of cell state. In this seminar, I will introduce Deep-Phase, a neural-network framework that converts condensate morphology into quantitative readouts of RNA biochemistry within them. By pairing automated fluorescence imaging with a convolutional classifier, we first resolve multiphase nucleolar states arising from small molecule perturbations of ribosome biogenesis and track their progression across treatment gradients. These morphology trajectories yield dose–response curves and morphology response concentrations that closely mirror biochemical potency values measured in orthogonal assays, establishing condensate shape as a sensitive fingerprint of underlying molecular activities. Applying the same approach to nuclear speckles and RSV inclusion bodies demonstrates its generality: morphology changes predict both splicing inhibitor potency and viral replication inhibition. Using Deep-Phase in a focused chemical screen also revealed an unexpected nucleolar phenotype—the “nucleolar flower”—and follow-up experiments uncovered a previously unrecognized role for a DNA topoisomerase in rRNA processing and the maintenance of multiphase nucleolar architecture. Together, Deep-Phase provides a scalable platform for quantitatively inferring condensate biochemistry from images alone, laying the foundation for future efforts to rationally engineer condensate form through targeted control of RNA activity as well as to discover morphology-guided biomarkers and modulators of RNA dysfunction in disease.
Bio:
Anita Ðonlić is a postdoctoral researcher at Princeton University and the Howard Hughes Medical Institute, where she integrates chemical, imaging, and AI-driven approaches to understand how biomolecular condensates modulate RNA function. She earned her Ph.D. in Chemistry from Duke University, where she pioneered small-molecule targeting of an oncogenic long non-coding RNA and contributed to building R-BIND, a foundational database for bioactive RNA-binding ligands. Her current work uses AI-guided quantitative microscopy to read out RNA biochemical states directly from condensate morphology, enabling the discovery of transient modulators of RNA function within these compartments. Building on these tools and insights, her future independent lab will uncover, engineer, and therapeutically modulate the molecular logic of RNA organization by condensates, opening new paths for understanding RNA function in cells and developing next-generation interventions.
Zoom link for attending remotely, if needed: https://epfl.zoom.us/j/69216732793
(DAY TWO: talk three / previous talk / next talk)
Abstract:
Biomolecular condensates serve as dynamic hubs for RNA-centered processes, including rRNA synthesis, pre-mRNA splicing, and viral RNA replication. However, linking condensate mesoscale morphology to the underlying biochemical activity of these RNA pathways has remained a fundamental challenge. Being able to infer RNA pathway disruptions directly from condensate imaging would enable quantitative decoding of condensate function, illuminate mechanisms of drug action, and transform microscopy data into predictive, functional readouts of cell state. In this seminar, I will introduce Deep-Phase, a neural-network framework that converts condensate morphology into quantitative readouts of RNA biochemistry within them. By pairing automated fluorescence imaging with a convolutional classifier, we first resolve multiphase nucleolar states arising from small molecule perturbations of ribosome biogenesis and track their progression across treatment gradients. These morphology trajectories yield dose–response curves and morphology response concentrations that closely mirror biochemical potency values measured in orthogonal assays, establishing condensate shape as a sensitive fingerprint of underlying molecular activities. Applying the same approach to nuclear speckles and RSV inclusion bodies demonstrates its generality: morphology changes predict both splicing inhibitor potency and viral replication inhibition. Using Deep-Phase in a focused chemical screen also revealed an unexpected nucleolar phenotype—the “nucleolar flower”—and follow-up experiments uncovered a previously unrecognized role for a DNA topoisomerase in rRNA processing and the maintenance of multiphase nucleolar architecture. Together, Deep-Phase provides a scalable platform for quantitatively inferring condensate biochemistry from images alone, laying the foundation for future efforts to rationally engineer condensate form through targeted control of RNA activity as well as to discover morphology-guided biomarkers and modulators of RNA dysfunction in disease.
Bio:
Anita Ðonlić is a postdoctoral researcher at Princeton University and the Howard Hughes Medical Institute, where she integrates chemical, imaging, and AI-driven approaches to understand how biomolecular condensates modulate RNA function. She earned her Ph.D. in Chemistry from Duke University, where she pioneered small-molecule targeting of an oncogenic long non-coding RNA and contributed to building R-BIND, a foundational database for bioactive RNA-binding ligands. Her current work uses AI-guided quantitative microscopy to read out RNA biochemical states directly from condensate morphology, enabling the discovery of transient modulators of RNA function within these compartments. Building on these tools and insights, her future independent lab will uncover, engineer, and therapeutically modulate the molecular logic of RNA organization by condensates, opening new paths for understanding RNA function in cells and developing next-generation interventions.
Zoom link for attending remotely, if needed: https://epfl.zoom.us/j/69216732793
Practical information
- Informed public
- Free
Organizer
- Prof. Matteo Dal Peraro, Institute of Bioengineering, School of Engineering, EPFL
Contact
- Institute of Bioengineering (IBI), Dietrich REINHARD