September 25
@
12:00 PM
–
1:00 PM
Learning Representations of Cellular Morphology
Juan Caicedo, PhD
Investigator, Morgridge Institute for Research
Assistant Professor, Department of Biostatistics and Medical Informatics
University of Wisconsin-Madison
Location Change: Union South Landmark A & B (3rd floor)
Abstract:
Microscopy images are fundamental for biological research, and quantifying cellular phenotypes is at the core of numerous applications in drug discovery, functional genomics and personalized medicine. However, quantifying cellular morphology can be a challenging problem because there is no universal reference to align the visual patterns observed under the microscope, and the structures of interest may also be unknown ahead of time. Deep learning offers a robust way to automatically identify and extract meaningful image-based representations, and specifically, self-supervised learning has the ability to discover biologically relevant structures in imaging data without prior knowledge or manual annotations. In this talk, I will show how image-based profiling powered by deep learning can support various biological studies at large scale, including phenotyping the impact of cancer variants, profiling subcellular protein localization, and predicting compound bioactivity.
Print PDF