bioconductor-cytomem

name

bioconductor-cytomem

description

MEM, Marker Enrichment Modeling, automatically generates and displays quantitative labels for cell populations that have been identified from single-cell data. The input for MEM is a dataset that has pre-clustered or pre-gated populations with cells in rows and features in columns. Labels convey a list of measured features and the features' levels of relative enrichment on each population. MEM can be applied to a wide variety of data types and can compare between MEM labels from flow cytometry,

cytoMEM

Workflows

Standard Workflow

Perform Marker Enrichment Modeling (MEM) analysis on single-cell data to generate quantitative population labels, visualize them via heatmaps, and calculate similarity scores between populations.

library(cytoMEM)
data(PBMC)

# Run MEM analysis on the PBMC dataset
MEM_values <- MEM(
  PBMC,
  transform = TRUE,
  cofactor = 15,
  choose.markers = FALSE,
  markers = "all",
  choose.ref = FALSE,
  zero.ref = FALSE,
  rename.markers = FALSE,
  new.marker.names = "none",
  IQR.thresh = NULL
)

# Generate population labels and heatmaps
build_heatmaps(
  MEM_values,
  cluster.MEM = "both",
  cluster.medians = "none",
  cluster.IQRs = "none",
  display.thresh = 1,
  output.files = FALSE,
  labels = FALSE,
  only.MEMheatmap = FALSE
)

# Calculate similarity (RMSD) scores between populations
MEM_RMSD(
  MEM_values[[5]][[1]],
  format = NULL,
  output.matrix = FALSE
)

Input/Output Note: Inputs a matrix or data frame of single-cell expression data with a cluster ID column; outputs MEM scores, enrichment labels, heatmaps, and population similarity matrices.

When to Use

• To automatically generate quantitative, human-readable enrichment labels for pre-clustered or pre-gated cell populations from single-cell data (e.g., mass cytometry or flow cytometry) using MEM().
• To visualize marker enrichment and population medians across clusters using build_heatmaps().
• To compare the similarity of MEM labels across different populations or datasets using root-mean-square deviation (RMSD) via MEM_RMSD().

When NOT to Use

• For clustering raw single-cell data from scratch; use packages like Seurat or scran because cytoMEM requires pre-clustered or pre-gated populations.
• For differential expression analysis of single-cell RNA-seq data; use edgeR or DESeq2 because cytoMEM is designed to generate descriptive enrichment labels rather than statistical hypothesis testing.
• For raw flow cytometry compensation or gating; use flowCore because cytoMEM expects pre-processed, gated, or clustered expression matrices.

Data Requirements

• Input format: Matrix, data frame, or file paths (.txt, .csv, .fcs).
• Structure: Cells in rows, markers/features in columns. The last column must be the cluster ID (unless file.is.clust = TRUE).
• Normalization/Transformation: Optionally transformed using hyperbolic arcsine (transform = TRUE with a specified cofactor, e.g., cofactor = 15 for mass cytometry).

Key Parameters

• transform (FALSE): Logical indicating whether to apply a hyperbolic arcsine transformation to the data.
• cofactor (15): The cofactor used in the hyperbolic arcsine transformation.
• choose.markers (FALSE): Logical indicating whether to select markers interactively via the console.
• markers ("all"): Character string specifying which markers to include in the analysis.
• choose.ref (FALSE): Logical indicating whether to choose an alternative reference population.
• zero.ref (FALSE): Logical indicating whether to use a zero (synthetic negative) reference.
• IQR.thresh (NULL): Threshold for interquartile range to avoid artificial inflation of MEM values (defaults internally to 0.5).
• display.thresh (0): Numeric value (0-10) specifying the minimum MEM score required for a marker to be displayed.

Best Practices

• Apply channel-specific transformations prior to running MEM() and set transform = FALSE if different cofactors are required for different fluorescence channels.
• Keep the default IQR.thresh of 0.5 unless you have a deep understanding of the dataset and the implications of changing it.
• Set labels = TRUE in build_heatmaps() to display the full MEM labels directly on the generated heatmap.
• Save the heatmap as a PDF file (output.files = TRUE) to prevent row names (MEM labels) from being cut off due to R window dimensions.

Common Pitfalls

• Row names cut off in heatmaps: Occurs due to R window dimensions. Fix: Set output.files = TRUE to save the heatmap as a PDF and open it in a PDF viewer, or read the full labels from the generated enrichment score-rownames.txt file.
• Artificial inflation of MEM values: Occurs when a population has a very small IQR due to background-level expression. Fix: Ensure IQR.thresh is set to at least 0.5 (the default) to threshold low IQR values.
• Mismatched columns across multiple files: Occurs when reading multiple files with different formats. Fix: Ensure all files have identical columns (features) and are of the same file type.

Alternatives

• Seurat: For comprehensive single-cell RNA-seq clustering and marker identification using differential expression.
• scater: For single-cell quality control and visualization of expression patterns.
• scran: For formal statistical testing of marker genes using differential expression.
• flowCore: For low-level flow cytometry data import, compensation, and transformation.

Citations

• Diggins, K. et al. (2017), Nature Methods.

References

• Homepage: bioconductor.org/packages/cytoMEM
• Vignette: https://bioconductor.org/packages/release/bioc/vignettes/cytoMEM/inst/doc/Intro_to_Marker_Enrichment_Modeling_Analysis.html