Cytofkit: A Bioconductor Package for an Integrated Mass Cytometry Data Analysis Pipeline.
Chen H, Lau MC, Wong MT, Newell EW, Poidinger M, Chen J. · 2016
Read Cytofkit: A Bioconductor Package for an Integrated Mass Cytometry Data Analysis Pipeline. by Chen H, Lau MC, Wong MT, Newell EW, Poidinger M, Chen J. (2016) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc5035035.txt.
Introduction
Hao Chen and six co-authors present a software package for mass cytometry in PLoS Computational Biology in 2016. Mass cytometry tags antibodies with metals instead of dyes, letting researchers measure more than forty markers on each single cell. The flood of high-dimensional data is hard to clean, view, and sort by hand. Their package, cytofkit, strings the whole analysis into one pipeline inside the free R-based Bioconductor system.
The pipeline handles pre-processing, shrinks the data into readable maps, finds cell subsets on its own, and works out how the subsets relate. A point-and-click window and an interactive app let bench scientists run it and explore the results without coding skill. On blood and tonsil test data, the tool recovered known lymphocyte groups and split helper T cells into finer subtypes across tissues. The same pipeline also works for standard flow cytometry data.
Catalog record
| CREATOR | Chen H, Lau MC, Wong MT, Newell EW, Poidinger M, Chen J. |
|---|---|
| YEAR | 2016 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 44 KB |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Cytofkit: A Bioconductor Package for an Integrated Mass Cytometry Data Analysis Pipeline By Hao Chen, Mai Chan Lau, Michael Thomas Wong, Evan W Newell, Michael Poidinger, Jinmiao Chen, Dina Schneidman Published in: PLoS Computational Biology DOI: 10.1371/journal.pcbi.1005112 ABSTRACT Abstract Single-cell mass cytometry significantly increases the dimensionality of cytometry analysis as compared to fluorescence flow cytometry, providing unprecedented resolution of cellular diversity in tissues. However, analysis and interpretation of these high-dimensional data poses a significant technical challenge. Here, we present cytofkit, a new Bioconductor package, which integrates both state-of-the-art bioinformatics methods and in-house novel algorithms to offer a comprehensive toolset for mass cytometry data analysis. Cytofkit provides functions for data pre-processing, data visualization through linear or non-linear dimensionality reduction, automatic identification of cell subsets, and inference of the relatedness between cell subsets. This pipeline also provides a graphical user interface (GUI) for ease of use, as well as a shiny application (APP) for interactive visualization of cell subpopulations and progression profiles of key markers. Applied to a CD14−CD19− PBMCs dataset, cytofkit accurately identified different subsets of lymphocytes; applied to a human CD4+ T cell dataset, cytofkit uncovered multiple subtypes of TFH cells spanning blood and tonsils. Cytofkit is implemented in R, licensed under the Artistic license 2.0, and freely available from the Bioconductor website, https://bioconductor.org/packages/cytofkit/. Cytofkit is also applicable for flow cytometry data analysis. "This is a PLOS Computational Biology Software Article." INTRODUCTION Mass cytometry, or cytometry by time-of-flight (CyTOF), uniquely combines metal-labeling of antibodies with mass spectrometry to enable high-dimensional measurement of the characteristics of individual cells [1,2]. The high purity and choice of metal isotopes overcome the limitations of spectral overlap in flow cytometry, and allow for simultaneous analysis of more than 40 markers per cell [3,4]. This technology has been successfully applied in a number of areas including mapping phenotypic heterogeneity of leukemia [5], inferring cellular progression and hierarchies [6], assessing drug effects on immune cells [7,8] and uncovering mechanisms of cellular reprogramming [9]. Despite the advantages of mass cytometry, effective analysis and interpretation of these high dimensional and large-scale datasets remain challenging. Traditional manual gating, the gold-standard method for flow cytometry data analysis, is not practical for mass cytometry due to its high dimensionality. In addition, most automated methods designed for flow cytometry data do not perform well for mass cytometry data [10]. Analysis of mass cytometry data has several key challenges including debarcoding [11], batch normalization [12], visualization of high-dimensional data, identification of cell subsets, inference of relatedness between cell subsets, and detection of changes in subset abundance. This manuscript focuses on addressing the following three key challenges for data that don’t display batch effect. The first challenge is efficient visualization of these high-dimensional data. A biaxial plot that displays the correlation of every two markers is a common way to visualize flow cytometry data. With the fact that m(m − 1)/2 biaxial plots are needed to fully visualize an m-dimensional dataset, this approach is impractical for mass cytometry data as the parameter m of mass cytometry is usually greater than 40. Alternative dimensionality reduction approaches have been used to transform the high-dimensional data to a low-dimensional representation, thus allowing visualization of the cells in a single plot. In Newell et al. [13], principal component analysis (PCA) was used to visualize a 25-parameter mass cytometry panel for CD8+ T cells. However, PCA is a linear transformation, and it cannot capture nonlinear relationships. To address this limitation, Amir et al. [5] developed a visualization tool named viSNE which utilizes the t-distributed stochastic neighbor embedding (t-SNE) algorithm. t-SNE is a nonlinear dimensionality reduction approach [14] which embeds the data from high dimensional space into a lower dimensional map based on similarities. On a t-SNE map, similar cells are placed to nearby points, while dissimilar cells are placed far apart. It has been demonstrated that t-SNE can effectively visualize cellular heterogeneity in normal and leukemic bone marrow [5]. The second challenge is to identify cell subpopulations. To address this challenge, the ACCENSE method has been developed to automatically identify cellular subpopulations using a density peak-finding algorithm on a t-SNE transformed 2-D map [15]. However, not all cells are assigned to a defined subpopulation in this method. DensVM extends ACCENSE by using support vector machine (SVM) to assign any unassigned cells to the subpopulations in a machine learning manner [16]. This approach has been demonstrated to precisely detect the boundaries of cell populations in murine myeloid data. DensVM has also been applied to map the numerous subtypes of follicular helper T cells derived from human blood and tonsils [16,17]. However, both ACCENSE and DensVM rely on a computationally intensive search for an optimal number of subpopulations. PhenoGraph, a graph-based partitioning method, has demonstrated efficiency in subpopulation detection [10]. PhenoGraph first constructs a nearest-neighbor graph which captures the phenotypic relatedness of the high-dimensional data, and then it applies a graph partition algorithm called Louvain [18] to dissect the nearest-neighbor graph into phenotypically coherent subpopulations. Applied to the study of acute myeloid leukemia, PhenoGraph provided a comprehensive view of the major phenotypes and elucidated intra- and inter-tumor heterogeneity. PhenoGraph has also been tested on three different mass cytometry datasets of healthy human bone marrow, and it displayed superior accuracy and robustness in immune cell type detection as compared to other methods. The third challenge is to detect cellular progression. In addition to defining distinctive cell subsets, there is great interest in resolving the order of cellular differentiation to reveal their developmental relationships. For example, Bendall et al. developed a graph-based trajectory detection algorithm named Wanderlust, which orders cells into a unified trajectory that reflects the developmental path [19]. This method correctly predicted the early developmental path of human B-lymphocytes. Nevertheless, this algorithm was designed for linear and non-branching developmental path, and hence is less useful for interpreting complex single-cell data with multiple developmental lineages. Wishbone extended the ability of Wanderlust to capture bifurcating developmental trajectories through introducing waypoints and identifying branch points [20]. Wishbone is based on diffusion map, which has been demonstrated to be powerful and robust for detecting the global geometric structures from the data [21]. However, Wishbone requires the input of a starting cell. SPADE is an innovative approach designed to extract cellular hierarchy using minimum spanning tree (MST) [6]. While SPADE enables the prediction of multi-branched cell developmental pathways, the hierarchical clustering used in SPADE needs a pre-specification of the number of clusters, additionally, the MST used by SPADE is susceptible to over-fitting and is not robust for local variation [9]. Our recent novel method named Mpath constructs multi-branching cell lineages from single-cell data using neighborhood-based cell state transitions [22]. However we have only demonstrated its applications for single-cell RNA-sequencing data. We are currently testing and optimizing Mpath for mass cytometry and flow cytometry data. In this report, we present an integrated analysis pipeline, named cytofkit. It is designed to analyze mass cytometry data in four main steps. In the first step, cytofkit performs data pre-processing, and enables combined analysis of multiple Flow Cytometry Standard (FCS) files. Users are allowed to customize their data merging strategy to combine the data using selectable transformation methods. The remaining three steps address respectively, each of the three challenges discussed above. Firstly,