1. A new interpretable machine learning approach for single-cell data discovers correlates of clinical outcome in cancer immunotherapy
- Author
-
Greg, Finak, Greene, Evan, D Amico, Leonard, Bhardwaj, Nina, Church, Candice, Morishima, Chihiro, Ramchurren, Nirasha, Taube, Janis, Paul Nghiem, Cheever, Martin, Fling, Steven, and Gottardo, Raphael
- Subjects
Immunology ,Immunology and Allergy - Abstract
High-dimensional cytometry is routinely used to characterize patient responses to cancer immunotherapy and other treatments, producing a wealth of datasets ripe for exploration but whose biological and technical heterogeneity make them difficult to analyze with current tools. We introduce a new interpretable machine learning method for single-cell mass and flow cytometry studies, FAUST, that robustly performs unbiased cell population discovery and annotation and enables data integration across studies and platforms. FAUST returns biologically interpretable cell phenotypes that can be compared across studies, making it well-suited for the analysis and integration of complex datasets. We use FAUST to perform candidate biomarker discovery and validation by applying it to flow and mass cytometry datasets from melanoma anti-PD-1 trials in order to discover and validate new CD4+ and CD8+ effector-memory T cell correlates of outcome co-expressing PD-1, HLA-DR, and CD28. Existing state-of-the-art computational discovery approaches as well as prior manual analyses did not detect any statistically significant T cell correlates associated with anti-PD-1 treatment these data. We use FAUST to replicate the discovery of published myeloid cell correlates and validate these by identifying them de novo in independent trials. FAUST’s phenotypic annotations are used to perform cross-study integration diverse data sets, allowing hypothesis-driven inference about cell sub-populations through a Phenotypic and Functional Differential Abundance (PFDA) framework. We showcase this approach on data across multiple trials. These results establish FAUST as a powerful new approach for unbiased discovery in single-cell cytometry.
- Published
- 2020
- Full Text
- View/download PDF