Skip to contents

Prepare Analysis Tracks on a SingleCellExperiment

Source: R/method-L1_analysis_tracks.R
prepare_sce.Rd

Runs the complete preprocessing pipeline for both feature spaces required for CellVoteR directly on the SCE object using standard Bioconductor functions.

Usage

prepare_sce(
  sce,
  marker_config,
  n_hvgs = 2000L,
  overlap_feat_percent = 50,
  n_pcs = NULL,
  k = NULL,
  resolution = NULL,
  cluster_params_args = list(),
  BPPARAM = BiocParallel::SerialParam()
)

Arguments

sce

A SingleCellExperiment with logcounts assay (from normalize_counts).

marker_config

Named list with $broad (from build_broad_marker_config) and $fine (from load_markers).

n_hvgs

Integer scalar. Number of HVGs. Defaults to 2000.

overlap_feat_percent

Numeric scalar (0-100). Minimum fine marker overlap. Defaults to 50.

n_pcs, k, resolution

Override automatic parameter estimation. NULL (default) uses estimate_cluster_params.

cluster_params_args

Named list passed to estimate_cluster_params.

BPPARAM

A BiocParallelParam. Defaults to SerialParam().

Value

The input SCE with both analysis tracks populated. The marker_config that is stored in the metadata of the altExp differs from that stored in the main SCE in that the fine marker sets are filtered to only include the genes present in the user panel and that overlap with the broad marker sets. The broad HVG track is built using the full logcounts assay and the full set of HVGs, while the user panel track is built using a subset of features based on the fine markers that overlap with the user panel and broad markers. The dimensionality reduction and clustering for each track are performed independently using their respective feature spaces, and their parameters are estimated separately using estimate_cluster_params.

Storage layout


sce
|-- assays: counts, logcounts
|-- rowSubset("broad_hvgs")
|-- reducedDim("PCA_broad_hvg")
|-- colData$cluster_broad_hvg
|-- metadata$broad_hvg_params
|-- metadata$marker_config
|-- metadata$filterd_fine_markers
|-- metadata$missing_by_label
`-- altExp("user_panel")
    |-- assays: counts, logcounts
    |-- reducedDim("PCA")
    |-- colData$cluster
    |-- metadata$marker_config
    |-- metadata$filterd_fine_markers
    `-- metadata$params

Examples

if (FALSE) { # \dontrun{
sce <- normalize_counts(sce)
sce <- prepare_sce(sce, marker_config)

# Broad HVG track
reducedDim(sce, "PCA_broad_hvg")
sce$cluster_broad_hvg

# User panel track
reducedDim(altExp(sce, "user_panel"), "PCA")
altExp(sce, "user_panel")$cluster
} # }