Beyond the Biaxial: Cytometry in the Era of High Dimensional Data and Machine Learning Analysis

The Speaker

Jonathan M. Irish, PhD - Professor | University of Colorado

Jonathan M. Irish Ph.D is Professor at the University of Colorado working in immunology, cancer, and computational biology and Scientific Director of the CytoLab data science group.  The Irish lab uses bench and computational cytometry techniques to study how signaling controls cell identity in healthy tissues, cancer, the human brain, and immune disorders.   Jonathan trained in chemistry and biology at the University of Michigan, went to Stanford for training in cancer biology and immunology, started his independent lab at Vanderbilt in systems cancer immunology, and was recruited to the University of Colorado in 2024 to focus on brain tumors and neuroimmunology.  A research theme in the Irish lab is the study of human cells and tissue using advanced cytometry, including phospho-flow, high dimensional mass cytometry and spectral flow, and machine learning analysis.  Jonathan is active in ISAC, including previously as Chair of Leadership Development and Data Committees and now as Secretary and Chair of Governance and an active member of the FlowRepository and FCS 4.0 taskforces.

Summary

Traditional cytometry analysis uses a series of biaxial gates to explore data and identify cells.  This approach works best when measured immunophenotypes match known cell types from hierarchical models of cell identity.  However, gating schemes may not accurately represent immune and cancer cell types that diverge from expected protein expression profiles or that exist in hybrid states that are between or apart from known cell types. I will present a new version of the Marker Enrichment Modeling algorithm, Velociraptor (MEM 4.0), which addresses these issues by quantifying cell identity using multidimensional, continuous measurements.  This approach enables flexible searching for cells based on feature sets expressed as readable text labels. This new approach also efficiently learns cell identity from training data and can seek any number of defined cell identities in new testing datasets, including cytometry from different instruments or platforms (e.g., training in imaging mass cytometry and validation in spectral flow cytometry). Scoring cell identity on a continuous scale is especially useful for characterizing cells that deviate from expected expression profiles. Such cells are commonly observed in human blood and tissue samples and are prevalent in disease and following activation of cell signaling. Additional applications of Velociraptor include measuring known cell subsets without manual gating, quantifying shifts in heterogeneity over time, and registering cells between blood and tissue microenvironments to track and characterize rare, clinically significant cells.

Learning Objectives:

1. Learn multiple approaches to identify cells in cytometry data
2. Understand the pros and cons of hierarchical, binary models of cell identity
3. Learn about disease-associated ‘iconoclast’ cells observed in human tissue

Who Should Attend:

- Trainees interested in high dimensional cytometry and associated data analysis
- Clinical researchers who want to identify rare, clinically significant cells in human tissue
- Immunologists and cell biologists who want to distinguish healthy and malignant cells
- Shared resource leaders who want to automate routine cell identification
- Data Scientists who want to integrate multiple single cell data types
-Hematopathologists and researchers who use biaxial gating to analyze flow data


Keywords: Cell identity, gating, cancer, immunology, machine learning

CMLE Credit: 1.0