An Unsupervised Learning Approach to Quantifying T Cell States From High-Plex Cytometry Data

→ Speaker: Dimitri Sidiropoulos, PhD Candidate, Johns Hopkins School of Medicine

Cytometric studies surveying cell phenotypes often do not account for functional states of immune cells that occur along continuous phenotypic transitions. To overcome this limitation, Sidiropoulos explains how he and his team have applied single-cell trajectory inference and non-negative matrix factorization methods to CyTOF® data to trace the dynamics of T cell states. In the setting of cancer immunotherapy, patient-specific summaries of continuous phenotypic shifts in T cells and estimation of patient-specific cell states can be inferred from peripheral blood-derived CyTOF data. The work establishes the utility of continuous metrics for CyTOF analysis as tools for translational biomarker discovery and provides a functional framework for immune system proteomics that can empower analysis in cancer immunotherapy.