5 June 2025, 10:30
Learning Cytometers: identifying critical cells beyond molecular markers
Abstract Traditional flow cytometry has relied heavily on fluorescent molecular markers and investigator driven hypotheses to detect and sort cells of interest However suitable molecular markers are often unavailable or insufficient to distinguish biologically critical cell subpopulations effectively To address this we have developed a series of data centric AI based instruments termed... Abstract[Traditional flow cytometry has relied heavily on fluorescent molecular markers and investigator-driven hypotheses to detect and sort cells of interest. However, suitable molecular markers are often unavailable or insufficient to distinguish biologically critical cell subpopulations effectively. To address this, we have developed a series of data-centric AI-based instruments, termed Learning Cytometer, to find and sort critical cells by leveraging high-content, label-free optical data. I will first introduce Ghost Cytometry (Science, 2018; eLife, 2021), which enabled the first and fast “imaging" cell sorter. By combining label-free and fluorescence image data analyzed through supervised and unsupervised machine learning approaches, Ghost Cytometry identifies cells based on complex morphological signatures, particularly beneficial when molecular markers are unavailable...
Speaker(s): Sadao Ota, RCAST at University of Tokyo,
ThinkCyte inc., Japan
Host: Beata Ramasz, Flow Cytometry Core Facility
Place: Small Operon
External Faculty Speaker
EMBL Heidelberg, Virtual
Additional information
Abstract
[Traditional flow cytometry has relied heavily on fluorescent molecular markers and investigator-driven hypotheses to detect and sort cells of interest. However, suitable molecular markers are often unavailable or insufficient to distinguish biologically critical cell subpopulations effectively. To address this, we have developed a series of data-centric AI-based instruments, termed Learning Cytometer, to find and sort critical cells by leveraging high-content, label-free optical data.
I will first introduce Ghost Cytometry (Science, 2018; eLife, 2021), which enabled the first and fast “imaging" cell sorter. By combining label-free and fluorescence image data analyzed through supervised and unsupervised machine learning approaches, Ghost Cytometry identifies cells based on complex morphological signatures, particularly beneficial when molecular markers are unavailable or undesirable. Using several case studies, I will show the capabilities and applications of Ghost Cytometry in classifying diverse cell phenotypes and responses, highlighting its potential as a tool for unbiased, high-content cell analysis and sorting in biomedical research.
Finally, as time permits, I will briefly introduce two recent advancements: Deep Nanometry (Nature Communications, 2025), a technology enabling ultrasensitive nanoparticle characterization, and another for time-lapse single-cell spectrometry (in submission).].
About the speaker
[Networked biophotonics and microfluidics
RCAST at Univ of Tokyo
ThinkCyte inc.].
Connection details
Zoom*: [https://embl-org.zoom.us/j/93633159006?pwd=AQjra4V7PEoZl5aEqp7bchjceOOSCZ.1] (Meeting ID: [93633159006]