Mass cytometry (CyTOF) allows for examination of dozens of proteins at single-cell resolution. By employing heavy metal isotopes rather than fluorescent tags, thereby significantly reducing spectral overlap, CyTOF enables generation of high-throughput high-dimensional cytometry data.
Given the emergence of replicated multi-condition experiments, a primary task in the analysis of any type of single-cell data is to make sample-level inferences, in order to identify i) differentially abundant subpopulations; and, ii) changes in expression at the subpopulation-level, i.e., differential states (DS), across conditions. Preceding such analyses, key challenges lie in data preprocessing (e.g., to remove artefactual signal), clustering (to define subpopulations), and dimension reduction.
In this talk, I will present a suite of tools for differential discovery in CyTOF data, including ‘CATALYST’ for preprocessing and visualization, ‘diffcyt’ for differential testing, and a comprehensive analysis pipeline that leverages R/Bioconductor infrastructure. Secondly, I will cover benchmarks of key analysis steps, such as clustering and dimension reduction. Finally, I will touch on how we transferred our DS analysis framework to scRNA-seq, and developed a complex, flexible simulation framework for method comparison, with the ‘muscat’ package.
In this workshop, we will cover an R-based pipeline for differential analysis of (replicated, multi-condition) high-dimensional mass cytometry data, which is largely based on Bioconductor infrastructure, and includes: i) identification of cell subpopulations using a sequence of high-resolution clustering, consensus clustering, manual merging and annotation; and, ii) differential abundance (DA) and state (DS) analyses, in order to identify association of population abundances with a phenotype, or changes in signalling within populations. Alongside formal statistical analyses, we will perform exploratory data analysis at each step, such as reporting on various clustering and differential testing results through dimensionality reduction, heatmaps of aggregated signal etc. *The workshop will closely follow Nowicka et al.’s “CyTOF workflow: differential discovery in high-throughput high-dimensional cytometry datasets” (F1000Research, 2017), available here.
Key words: mass cytometry; CyTOF; visualization; clustering; dimension reduction; differential analysis
Technical: You will need to bring your own laptop. The workshop will use cloud-based resources, so your laptop will need a web browser and WiFi capabilities.
Knowledge/competencies: Participants are expected to have basic-intermediate knowledge of R and some familiarity with Bioconductor’s SingleCellExperiment class.
Relevance: The workshop presented here will equip participants with the expertise for diverse exploratory and differential analyses of high-dimensional cytometry data with complex experimental design, i.e., multiple cell subpopulations, samples (e.g. patients), and conditions (e.g. treatments). Furthermore, a large proportion of the analyses presented here are transferable to scRNA-seq, and the workshop may thus be of interest also to anyone who is interested in analysing replicated multi-condition scRNA-seq data.
Institute of Molecular Life Sciences, University of Zurich
Helena earned her undergraduate degree at the Univeristy of Heidelberg in Biochemistry. She then went on to earn her Master’s degree in Computational Biology & Bioinformatics at the ETH Zurich. She is currently a PhD candidate in Statistical Bioinformatics at the University of Zurich.
Helena focuses on developing analysis frameworks for CyTOF data and differential discovery in scRNA-seq data. She is the author of a popular Bioconductor package providing tools for preprocessing and analysis of cytometry data.