Levine_32dim_SE: 'Levine_32dim' dataset

In lmweber/HDCytoData: Collection of high-dimensional cytometry benchmark datasets in Bioconductor object formats

'Levine_32dim' dataset

Description

Mass cytometry (CyTOF) dataset from Levine et al. (2015), containing 32 dimensions (surface protein markers). Manually gated cell population labels are available for 14 populations. Cells are human bone marrow cells from 2 healthy donors. This dataset can be used to benchmark clustering algorithms.

Usage

Levine_32dim_SE(metadata = FALSE)
Levine_32dim_flowSet(metadata = FALSE)

Arguments

metadata

logical value indicating whether ExperimentHub metadata (describing the overall dataset) should be returned only, or if the whole dataset should be loaded. Default = FALSE, which loads the whole dataset.

Details

This is a 32-dimensional mass cytometry (CyTOF) data set, consisting of expression levels of 32 surface marker proteins. Cell population labels are available for 14 manually gated populations. Cells are human bone marrow cells from 2 healthy donors. Manually gated cell population labels were provided by the original authors.

This dataset can be used to benchmark clustering algorithms.

The dataset contains cells from 2 patients ('H1' and 'H2'); a total of 265,627 cells (104,184 manually gated and 161,443 unclassified); 14 manually gated cell population IDs (as well as 'unassigned'); and a total of 32 surface marker proteins.

The dataset is provided in two Bioconductor object formats: SummarizedExperiment and flowSet. In each case, cells are stored in rows, and protein markers in columns (this is the usual format used for flow and mass cytometry data).

For the link{SummarizedExperiment}, row and column metadata can be accessed with the rowData and colData accessor functions from the SummarizedExperiment package. The row data contains patient IDs and manually gated cell population IDs. The column data contains channel names, protein marker names, and a factor marker_class to identify the class of each protein marker ('cell type', 'cell state', as well as 'none' for any non protein marker columns that are not needed for downstream analyses; for this dataset, all proteins are cell type markers). The expression values for each cell can be accessed with assay. The expression values are formatted as a single table.

For the flowSet, the expression values are stored in a separate table for each sample. Each sample is represented by one flowFrame object within the overall flowSet. The expression values can be accessed with the exprs function from the flowCore package. Row metadata is stored as additional columns of data within the flowFrame for each sample; note that factor values are converted to numeric values, since the expression tables must be numeric matrices. Channel names are stored in the column names of the expression tables. Additional row and column metadata is stored in the description slots, which can be accessed with the description accessor function for the individual flowFrames; this includes additional sample information (where available), marker information, and cell population information.

Prior to performing any downstream analyses, the expression values should be transformed. A standard transformation used for mass cytometry data is the asinh with cofactor = 5.

File sizes: 44.2 MB (SummarizedExperiment and flowSet).

Original source: "benchmark data set 2" in Levine et al. (2015): https://www.ncbi.nlm.nih.gov/pubmed/26095251

Original link to raw data: https://www.cytobank.org/cytobank/experiments/46102 (download the .zip file shown under "Exported Files")

This dataset was previously used to benchmark clustering algorithms for high-dimensional cytometry in our article, Weber and Robinson (2016): https://www.ncbi.nlm.nih.gov/pubmed/27992111

Data files are also available from FlowRepository (FR-FCM-ZZPH): http://flowrepository.org/id/FR-FCM-ZZPH

Value

Returns a SummarizedExperiment or flowSet object.

References

Levine et al. (2015), "Data-driven phenotypic dissection of AML reveals progenitor-like cells that correlate with prognosis", Cell, 162, 184-197: https://www.ncbi.nlm.nih.gov/pubmed/26095251

Weber and Robinson (2016), "Comparison of clustering methods for high-dimensional single-cell flow and mass cytometry data", Cytometry Part A, 89A, 1084-1096: https://www.ncbi.nlm.nih.gov/pubmed/27992111

Examples

Levine_32dim_SE()
Levine_32dim_flowSet()

lmweber/HDCytoData documentation built on March 19, 2024, 4:41 a.m.