0.1 Introduction

CITE-seq data provide RNA and surface protein counts for the same cells. This tutorial shows how MuData can be integrated into with Bioconductor workflows to analyse CITE-seq data.

0.2 Installation

The most recent dev build can be installed from GitHub:


Stable version of MuData will be available in future bioconductor versions.

0.3 Loading libraries

#> Warning: replacing previous import 'utils::findMatches' by
#> 'S4Vectors::findMatches' when loading 'AnnotationDbi'


0.4 Loading data

We will use CITE-seq data accessible with the SingleCellMultiModal Bioconductor package, which was originally described in Stoeckius et al., 2017.

mae <- CITEseq(
    DataType="cord_blood", modes="*", dry.run=FALSE, version="1.0.0"
#> Dataset: cord_blood
#> Working on: scADT_Counts
#> Working on: scRNAseq_Counts
#> see ?SingleCellMultiModal and browseVignettes('SingleCellMultiModal') for documentation
#> loading from cache
#> see ?SingleCellMultiModal and browseVignettes('SingleCellMultiModal') for documentation
#> loading from cache

#> A MultiAssayExperiment object of 2 listed
#>  experiments with user-defined names and respective classes.
#>  Containing an ExperimentList class object of length 2:
#>  [1] scADT: matrix with 13 rows and 8617 columns
#>  [2] scRNAseq: matrix with 36280 rows and 8617 columns
#> Functionality:
#>  experiments() - obtain the ExperimentList instance
#>  colData() - the primary/phenotype DataFrame
#>  sampleMap() - the sample coordination DataFrame
#>  `$`, `[`, `[[` - extract colData columns, subset, or experiment
#>  *Format() - convert into a long or wide DataFrame
#>  assays() - convert ExperimentList to a SimpleList of matrices
#>  exportClass() - save data to flat files

We see two modalities in the object — scRNAseq and scADT, the latter providing counts for antibody-derived tags. Notably, each experiment is a matrix.

0.5 Processing ADT data

While CITE-seq analysis workflows such as CiteFuse should be consulted for more details, below we exemplify simple data transformation in order to demonstrate how their output can be saved to an H5MU file later on.

For ADT counts, we will apply CLR transformation following Hao et al., 2020:

# Define CLR transformation as in the Seurat workflow
clr <- function(data) t(
  apply(data, 1, function(x) log1p(
    x / (exp(sum(log1p(x[x > 0]), na.rm = TRUE) / length(x)))

We will make the ADT modality a SingleCellExperiment object and add an assay with CLR-transformed counts:

adt_counts <- mae[["scADT"]]

mae[["scADT"]] <- SingleCellExperiment(adt_counts)
assay(mae[["scADT"]], "clr") <- clr(adt_counts)

We will also generate reduced dimensions taking advantage of the functionality in the scater package:

mae[["scADT"]] <- runPCA(
  mae[["scADT"]], exprs_values = "clr", ncomponents = 20
#> Warning in check_numbers(k = k, nu = nu, nv = nv, limit = min(dim(x)) - : more
#> singular values/vectors requested than available
#> Warning in (function (A, nv = 5, nu = nv, maxit = 1000, work = nv + 7, reorth =
#> TRUE, : You're computing too large a percentage of total singular values, use a
#> standard svd instead.
plotReducedDim(mae[["scADT"]], dimred = "PCA",
               by_exprs_values = "clr", colour_by = "CD3")