Package 'rliger'

Description

Generate dot plot from input matrix with ComplexHeatmap

Usage

.complexHeatmapDotPlot(
  colorMat,
  sizeMat,
  featureAnnDF = NULL,
  cellSplitVar = NULL,
  cellLabels = NULL,
  maxDotSize = 4,
  clusterFeature = FALSE,
  clusterCell = FALSE,
  legendColorTitle = "Matrix Value",
  legendSizeTitle = "Fraction Value",
  transpose = FALSE,
  baseSize = 8,
  cellTextSize = NULL,
  featureTextSize = NULL,
  cellTitleSize = NULL,
  featureTitleSize = NULL,
  legendTextSize = NULL,
  legendTitleSize = NULL,
  featureGrpRot = 0,
  viridisOption = "C",
  viridisDirection = -1,
  ...
)

Arguments

Value

A HeatmapList object.

Description

Produce single violin plot with data frame passed from upstream

Usage

.ggCellViolin(
  plotDF,
  y,
  groupBy = NULL,
  colorBy = NULL,
  violin = TRUE,
  violinAlpha = 0.8,
  violinWidth = 0.9,
  box = FALSE,
  boxAlpha = 0.6,
  boxWidth = 0.4,
  dot = FALSE,
  dotColor = "black",
  dotSize = getOption("ligerDotSize"),
  raster = NULL,
  seed = 1,
  ...
)

Arguments

Value

ggplot object by default. When plotly = TRUE, returns plotly (htmlwidget) object.

Description

Controls content and size of all peripheral texts.

Usage

.ggplotLigerTheme(
  plot,
  title = NULL,
  subtitle = NULL,
  xlab = TRUE,
  ylab = TRUE,
  legendColorTitle = NULL,
  legendFillTitle = NULL,
  legendShapeTitle = NULL,
  legendSizeTitle = NULL,
  showLegend = TRUE,
  legendPosition = "right",
  baseSize = getOption("ligerBaseSize"),
  titleSize = NULL,
  subtitleSize = NULL,
  xTextSize = NULL,
  xFacetSize = NULL,
  xTitleSize = NULL,
  yTextSize = NULL,
  yFacetSize = NULL,
  yTitleSize = NULL,
  legendTextSize = NULL,
  legendTitleSize = NULL,
  legendDotSize = 4,
  panelBorder = FALSE,
  legendNRow = NULL,
  legendNCol = NULL,
  colorLabels = NULL,
  colorValues = NULL,
  colorPalette = "magma",
  colorDirection = -1,
  naColor = "#DEDEDE",
  colorLow = NULL,
  colorMid = NULL,
  colorHigh = NULL,
  colorMidPoint = NULL,
  plotly = FALSE
)

Arguments

Value

Updated ggplot object by default. When plotly = TRUE, returns plotly (htmlwidget) object.

Description

Produce single scatter plot with data frame passed from upstream

Usage

.ggScatter(
  plotDF,
  x,
  y,
  colorBy = NULL,
  shapeBy = NULL,
  dotOrder = c("shuffle", "ascending", "descending"),
  dotSize = getOption("ligerDotSize"),
  dotAlpha = 0.9,
  trimHigh = NULL,
  trimLow = NULL,
  zeroAsNA = TRUE,
  raster = NULL,
  labelBy = colorBy,
  labelText = TRUE,
  labelTextSize = 4,
  seed = 1,
  ...
)

Arguments

Details

Having package "ggrepel" installed can help adding tidier text labels on the scatter plot.

Value

ggplot object by default. When plotly = TRUE, returns plotly (htmlwidget) object.

Description

This is not an exported function. This documentation just serves for a manual of extra arguments that users can use when generating heatmaps with plotGeneHeatmap or plotFactorHeatmap.

Note that the following arguments are pre-occupied by upstream wrappers so users should not include them in a function call: dataMatrix, dataName, cellDF, featureDF, cellSplitVar, featureSplitVar.

The following arguments of Heatmap is occupied by this function, so users should include them in a function call as well: matrix, name, col, heatmap_legend_param, top_annotation, column_title_gp, column_names_gp, show_column_names, column_split, column_gap, left_annotation, row_title_gp, row_names_gp, show_row_names, row_split, row_gap.

Usage

.plotHeatmap(
  dataMatrix,
  dataName = "Value",
  cellDF = NULL,
  featureDF = NULL,
  transpose = FALSE,
  cellSplitVar = NULL,
  featureSplitVar = NULL,
  dataScaleFunc = NULL,
  showCellLabel = FALSE,
  showCellLegend = TRUE,
  showFeatureLabel = TRUE,
  showFeatureLegend = TRUE,
  cellAnnColList = NULL,
  featureAnnColList = NULL,
  scale = FALSE,
  trim = c(-2, 2),
  baseSize = 8,
  cellTextSize = NULL,
  featureTextSize = NULL,
  cellTitleSize = NULL,
  featureTitleSize = NULL,
  legendTextSize = NULL,
  legendTitleSize = NULL,
  viridisOption = "A",
  viridisDirection = -1,
  RColorBrewerOption = "RdBu",
  ...
)

Arguments

Value

HeatmapList-class object

Description

This function converts data stored in SingleCellExperiment (SCE), Seurat object or a merged sparse matrix (dgCMatrix) into a liger object. This is designed for a container object or matrix that already contains multiple datasets to be integerated with LIGER. For individual datasets, please use createLiger instead.

Usage

## S3 method for class 'dgCMatrix'
as.liger(object, datasetVar = NULL, modal = NULL, ...)

## S3 method for class 'SingleCellExperiment'
as.liger(object, datasetVar = NULL, modal = NULL, ...)

## S3 method for class 'Seurat'
as.liger(object, datasetVar = NULL, modal = NULL, assay = NULL, ...)

seuratToLiger(object, datasetVar = NULL, modal = NULL, assay = NULL, ...)

as.liger(object, ...)

Arguments

Details

For Seurat V5 structure, it is highly recommended that users make use of its split layer feature, where things like "counts", "data", and "scale.data" can be held for each dataset in the same Seurat object, e.g. with "count.ctrl", "count.stim", not merged. If a Seurat object with split layers is given, datasetVar will be ignored and the layers will be directly used.

Value

a liger object.

Examples

# dgCMatrix (common sparse matrix class), usually obtained from other
# container object, and contains multiple samples merged in one.
matList <- rawData(pbmc)
multiSampleMatrix <- mergeSparseAll(matList)
# The `datasetVar` argument expects the variable assigning the sample source
pbmc2 <- as.liger(multiSampleMatrix, datasetVar = pbmc$dataset)
pbmc2


if (requireNamespace("SingleCellExperiment", quietly = TRUE)) {
    sce <- SingleCellExperiment::SingleCellExperiment(
        assays = list(counts = multiSampleMatrix)
    )
    sce$sample <- pbmc$dataset
    pbmc3 <- as.liger(sce, datasetVar = "sample")
    pbmc3
}

if (requireNamespace("Seurat", quietly = TRUE)) {
    seu <- SeuratObject::CreateSeuratObject(multiSampleMatrix)
    # Seurat creates variable "orig.ident" by identifying the cell barcode
    # prefixes, which is indeed what we need in this case. Users might need
    # to be careful and have it confirmed first.
    pbmc4 <- as.liger(seu, datasetVar = "orig.ident")
    pbmc4

    # As per Seurat V5 updates with layered data, specifically helpful udner the
    # scenario of dataset integration. "counts" and etc for each datasets can be
    # split into layers.
    seu5 <- seu
    seu5[["RNA"]] <- split(seu5[["RNA"]], pbmc$dataset)
    print(SeuratObject::Layers(seu5))
    pbmc5 <- as.liger(seu5)
    pbmc5
}

Description

Works for converting a matrix or container object to a single ligerDataset, and can also convert the modality preset of a ligerDataset. When used with a dense matrix object, it automatically converts the matrix to sparse form (dgCMatrix-class). When used with container objects such as Seurat or SingleCellExperiment, it is highly recommended that the object contains only one dataset/sample which is going to be integrated with LIGER. For multi-sample objects, please use as.liger with dataset source variable specified.

Usage

## S3 method for class 'ligerDataset'
as.ligerDataset(
  object,
  modal = c("default", "rna", "atac", "spatial", "meth"),
  ...
)

## Default S3 method:
as.ligerDataset(
  object,
  modal = c("default", "rna", "atac", "spatial", "meth"),
  ...
)

## S3 method for class 'matrix'
as.ligerDataset(
  object,
  modal = c("default", "rna", "atac", "spatial", "meth"),
  ...
)

## S3 method for class 'Seurat'
as.ligerDataset(
  object,
  modal = c("default", "rna", "atac", "spatial", "meth"),
  assay = NULL,
  ...
)

## S3 method for class 'SingleCellExperiment'
as.ligerDataset(
  object,
  modal = c("default", "rna", "atac", "spatial", "meth"),
  ...
)

as.ligerDataset(object, ...)

Arguments

Value

a liger object.

Examples

ctrl <- dataset(pbmc, "ctrl")
ctrl
# Convert the modality preset
as.ligerDataset(ctrl, modal = "atac")
rawCounts <- rawData(ctrl)
class(rawCounts)
as.ligerDataset(rawCounts)

Description

liger object of bone marrow subsample data with RNA and ATAC modality

Usage

bmmc

Format

liger object with two dataset named by "rna" and "atac"

Source

https://www.nature.com/articles/s41587-019-0332-7

References

Jeffrey M. Granja and et. al., Nature Biotechnology, 2019

Description

This metric quantifies how much the factorization and alignment distorts the geometry of the original datasets. The greater the agreement, the less distortion of geometry there is. This is calculated by performing dimensionality reduction on the original and quantile aligned (or just factorized) datasets, and measuring similarity between the k nearest neighbors for each cell in original and aligned datasets. The Jaccard index is used to quantify similarity, and is the final metric averages across all cells.

Note that for most datasets, the greater the chosen nNeighbor, the greater the agreement in general. Although agreement can theoretically approach 1, in practice it is usually no higher than 0.2-0.3.

Usage

calcAgreement(
  object,
  ndims = 40,
  nNeighbors = 15,
  useRaw = FALSE,
  byDataset = FALSE,
  seed = 1,
  dr.method = NULL,
  k = nNeighbors,
  use.aligned = NULL,
  rand.seed = seed,
  by.dataset = byDataset
)

Arguments

Value

A numeric vector of agreement metric. A single value if byDataset = FALSE or each dataset a value otherwise.

Examples

if (requireNamespace("RcppPlanc", quietly = TRUE)) {
    pbmc <- pbmc %>%
    normalize %>%
    selectGenes %>%
    scaleNotCenter %>%
    runINMF %>%
    quantileNorm
    calcAgreement(pbmc)
}

Description

This metric quantifies how well-aligned two or more datasets are. We randomly downsample all datasets to have as many cells as the smallest one. We construct a nearest-neighbor graph and calculate for each cell how many of its neighbors are from the same dataset. We average across all cells and compare to the expected value for perfectly mixed datasets, and scale the value from 0 to 1. Note that in practice, alignment can be greater than 1 occasionally.

Usage

calcAlignment(
  object,
  clustersUse = NULL,
  clusterVar = NULL,
  nNeighbors = NULL,
  cellIdx = NULL,
  cellComp = NULL,
  resultBy = c("all", "dataset", "cell"),
  seed = 1,
  k = nNeighbors,
  rand.seed = seed,
  cells.use = cellIdx,
  cells.comp = cellComp,
  clusters.use = clustersUse,
  by.cell = NULL,
  by.dataset = NULL
)

Arguments

Details

$\bar{x}$ is the average number of neighbors belonging to any cells' same dataset, $N$ is the number of datasets, $k$ is the number of neighbors in the KNN graph.

$1 - \frac{\bar{x} - \frac{k}{N}}{k - \frac{k}{N}}$

The selection on cells to be measured can be done in various way and represent different scenarios:

1. By default, all cells are considered and the alignment across all datasets will be calculated.

2. Select clustersUse from clusterVar to use cells from the clusters of interests. This measures the alignment across all covered datasets within the specified clusters.

3. Only Specify cellIdx for flexible selection. This measures the alignment across all covered datasets within the specified cells. A none-NULL cellIdx privileges over clustersUse.

4. Specify cellIdx and cellComp at the same time, so that the original dataset source will be ignored and cells specified by each argument will be regarded as from each a dataset. This measures the alignment between cells specified by the two arguments. cellComp can contain cells already specified in cellIdx.

Value

The alignment metric.

Examples

if (requireNamespace("RcppPlanc", quietly = TRUE)) {
    pbmc <- pbmc %>%
    normalize %>%
    selectGenes %>%
    scaleNotCenter %>%
    runINMF %>%
    quantileNorm
    calcAlignment(pbmc)
}

Description

This function aims at calculating the adjusted Rand index for the clustering result obtained with LIGER and the external clustering (existing "true" annotation). ARI ranges from 0 to 1, with a score of 0 indicating no agreement between clusterings and 1 indicating perfect agreement.

The true clustering annotation must be specified as the base line. We suggest setting it to the object cellMeta so that it can be easily used for many other visualization and evaluation functions.

The ARI can be calculated for only specified datasets, since true annotation might not be available for all datasets. Evaluation for only one or a few datasets can be done by specifying useDatasets. If useDatasets is specified, the argument checking for trueCluster and useCluster will be enforced to match the cells in the specified datasets.

Usage

calcARI(
  object,
  trueCluster,
  useCluster = NULL,
  useDatasets = NULL,
  verbose = getOption("ligerVerbose", TRUE),
  classes.compare = trueCluster
)

Arguments

Value

A numeric scalar, the ARI of the clustering result indicated by useCluster compared to trueCluster.

References

L. Hubert and P. Arabie (1985) Comparing Partitions, Journal of the Classification, 2, pp. 193-218.

Examples

# Assume the true cluster in `pbmcPlot` is "leiden_cluster"
# generate fake new labeling
fake <- sample(1:7, ncol(pbmcPlot), replace = TRUE)
# Insert into cellMeta
pbmcPlot$new <- factor(fake)
calcARI(pbmcPlot, trueCluster = "leiden_cluster", useCluster = "new")

# Now assume we got existing base line annotation only for "stim" dataset
nStim <- ncol(dataset(pbmcPlot, "stim"))
stimTrueLabel <- factor(fake[1:nStim])
# Insert into cellMeta
cellMeta(pbmcPlot, "stim_true_label", useDatasets = "stim") <- stimTrueLabel
# Assume "leiden_cluster" is the clustering result we got and need to be
# evaluated
calcARI(pbmcPlot, trueCluster = "stim_true_label",
        useCluster = "leiden_cluster", useDatasets = "stim")

Description

This score represents the relative magnitude of the dataset-specific components of each factor's gene loadings compared to the shared components for two datasets. First, for each dataset we calculate the norm of the sum of each factor's shared loadings ($W$) and dataset-specific loadings ($V$). We then determine the ratio of these two values and subtract from 1... TODO: finish description.

Usage

calcDatasetSpecificity(
  object,
  dataset1,
  dataset2,
  doPlot = FALSE,
  do.plot = doPlot
)

Arguments

Value

List containing three elements.

Description

This function aims at calculating the purity for the clustering result obtained with LIGER and the external clustering (existing "true" annotation). Purity can sometimes be a more useful metric when the clustering to be tested contains more subgroups or clusters than the true clusters. Purity ranges from 0 to 1, with a score of 1 representing a pure, accurate clustering.

The true clustering annotation must be specified as the base line. We suggest setting it to the object cellMeta so that it can be easily used for many other visualization and evaluation functions.

The purity can be calculated for only specified datasets, since true annotation might not be available for all datasets. Evaluation for only one or a few datasets can be done by specifying useDatasets. If useDatasets is specified, the argument checking for trueCluster and useCluster will be enforced to match the cells in the specified datasets.

Usage

calcPurity(
  object,
  trueCluster,
  useCluster = NULL,
  useDatasets = NULL,
  verbose = getOption("ligerVerbose", TRUE),
  classes.compare = trueCluster
)

Arguments

Value

A numeric scalar, the purity of the clustering result indicated by useCluster compared to trueCluster.

Examples

# Assume the true cluster in `pbmcPlot` is "leiden_cluster"
# generate fake new labeling
fake <- sample(1:7, ncol(pbmcPlot), replace = TRUE)
# Insert into cellMeta
pbmcPlot$new <- factor(fake)
calcPurity(pbmcPlot, trueCluster = "leiden_cluster", useCluster = "new")

# Now assume we got existing base line annotation only for "stim" dataset
nStim <- ncol(dataset(pbmcPlot, "stim"))
stimTrueLabel <- factor(fake[1:nStim])
# Insert into cellMeta
cellMeta(pbmcPlot, "stim_true_label", useDatasets = "stim") <- stimTrueLabel
# Assume "leiden_cluster" is the clustering result we got and need to be
# evaluated
calcPurity(pbmcPlot, trueCluster = "stim_true_label",
           useCluster = "leiden_cluster", useDatasets = "stim")

Description

When need to interact with the data embedded in HDF5 files out of the currect R session, the HDF5 files has to be closed in order to be available to other processes.

Usage

closeAllH5(object)

## S3 method for class 'liger'
closeAllH5(object)

## S3 method for class 'ligerDataset'
closeAllH5(object)

Arguments

Value

Nothing is returned.

Description

Check difference of two liger command

Usage

commandDiff(object, cmd1, cmd2)

Arguments

Value

If any difference found, a character vector summarizing all differences

Examples

pbmc <- normalize(pbmc)
pbmc <- normalize(pbmc, log = TRUE, scaleFactor = 1e4)
cmds <- commands(pbmc)
commandDiff(pbmc, cmds[1], cmds[2])

Description

Convert old liger object to latest version

Usage

convertOldLiger(
  object,
  dimredName,
  clusterName = "clusters",
  h5FilePath = NULL
)

Arguments

Examples

## Not run: 
# Suppose you have a liger object of old version (<1.99.0)
newLig <- convertOldLiger(oldLig)

## End(Not run)

Description

Similar as how default ligerDataset data is accessed.

Usage

coordinate(x, dataset)

coordinate(x, dataset, check = TRUE) <- value

## S4 method for signature 'liger,character'
coordinate(x, dataset)

## S4 replacement method for signature 'liger,character'
coordinate(x, dataset, check = TRUE) <- value

## S4 method for signature 'ligerSpatialDataset,missing'
coordinate(x, dataset = NULL)

## S4 replacement method for signature 'ligerSpatialDataset,missing'
coordinate(x, dataset = NULL, check = TRUE) <- value

Arguments

Value

The retrieved coordinate matrix or the updated x object.

Description

For convenience, the default formatType = "10x" directly fits the structure of cellranger output. formatType = "anndata" works for current AnnData H5AD file specification (see Details). If a customized H5 file structure is presented, any of the rawData, indicesName, indptrName, genesName, barcodesName should be specified accordingly to override the formatType preset.

DO make a copy of the H5AD files because rliger functions write to the files and they will not be able to be read back to Python. This will be fixed in the future.

Usage

createH5LigerDataset(
  h5file,
  formatType = "10x",
  rawData = NULL,
  normData = NULL,
  scaleData = NULL,
  barcodesName = NULL,
  genesName = NULL,
  indicesName = NULL,
  indptrName = NULL,
  anndataX = "X",
  modal = c("default", "rna", "atac", "spatial", "meth"),
  featureMeta = NULL,
  ...
)

Arguments

Details

For H5AD file written from an AnnData object, we allow using formatType = "anndata" for the function to infer the proper structure. However, while a typical AnnData-based analysis tends to in-place update the adata.X attribute and there is no standard/forced convention for where the raw count data, as needed from LIGER, is stored. Therefore, we expose argument anndataX for specifying this information. The default value "X" looks for adata.X. If the raw data is stored in a layer, e.g. adata.layers['count'], then anndataX = "layers/count". If it is stored to adata.raw.X, then anndataX = "raw/X". If your AnnData object does not have the raw count retained, you will have to go back to the Python work flow to have it inserted at desired object space and re-write the H5AD file, or just go from upstream source files with which the AnnData was originally created.

Value

H5-based ligerDataset object

Examples

h5Path <- system.file("extdata/ctrl.h5", package = "rliger")
tempPath <- tempfile(fileext = ".h5")
file.copy(from = h5Path, to = tempPath)
ld <- createH5LigerDataset(tempPath)

Description

This function allows creating liger object from multiple datasets of various forms (See rawData).

DO make a copy of the H5AD files because rliger functions write to the files and they will not be able to be read back to Python. This will be fixed in the future.

Usage

createLiger(
  rawData,
  modal = NULL,
  cellMeta = NULL,
  removeMissing = TRUE,
  addPrefix = "auto",
  formatType = "10X",
  anndataX = "X",
  dataName = NULL,
  indicesName = NULL,
  indptrName = NULL,
  genesName = NULL,
  barcodesName = NULL,
  newH5 = TRUE,
  verbose = getOption("ligerVerbose", TRUE),
  ...,
  raw.data = rawData,
  take.gene.union = NULL,
  remove.missing = removeMissing,
  format.type = formatType,
  data.name = dataName,
  indices.name = indicesName,
  indptr.name = indptrName,
  genes.name = genesName,
  barcodes.name = barcodesName
)

Arguments

See Also

createLigerDataset, createH5LigerDataset

Examples

# Create from raw count matrices
ctrl.raw <- rawData(pbmc, "ctrl")
stim.raw <- rawData(pbmc, "stim")
pbmc1 <- createLiger(list(ctrl = ctrl.raw, stim = stim.raw))

# Create from H5 files
h5Path <- system.file("extdata/ctrl.h5", package = "rliger")
tempPath <- tempfile(fileext = ".h5")
file.copy(from = h5Path, to = tempPath)
lig <- createLiger(list(ctrl = tempPath))

# Create from other container object
if (requireNamespace("SeuratObject", quietly = TRUE)) {
    ctrl.seu <- SeuratObject::CreateSeuratObject(ctrl.raw)
    stim.seu <- SeuratObject::CreateSeuratObject(stim.raw)
    pbmc2 <- createLiger(list(ctrl = ctrl.seu, stim = stim.seu))
}

Description

Create in-memory ligerDataset object

Usage

createLigerDataset(
  rawData = NULL,
  modal = c("default", "rna", "atac", "spatial", "meth"),
  normData = NULL,
  scaleData = NULL,
  featureMeta = NULL,
  ...
)

Arguments

See Also

ligerDataset, ligerATACDataset, ligerSpatialDataset, ligerMethDataset

Examples

ctrl.raw <- rawData(pbmc, "ctrl")
ctrl.ld <- createLigerDataset(ctrl.raw)

Description

This function mainly aims at downsampling datasets to a size suitable for plotting or expensive in-memmory calculation.

Users can balance the sample size of categories of interests with balance. Multi-variable specification to balance is supported, so that at most maxCells cells will be sampled from each combination of categories from the variables. For example, when two datasets are presented and three clusters labeled across them, there would then be at most $2 \times 3 \times maxCells$ cells being selected. Note that "dataset" will automatically be added as one variable when balancing the downsampling. However, if users want to balance the downsampling solely basing on dataset origin, users have to explicitly set balance = "dataset".

Usage

downsample(
  object,
  balance = NULL,
  maxCells = 1000,
  useDatasets = NULL,
  seed = 1,
  returnIndex = FALSE,
  ...
)

Arguments

Value

By default, a subset of liger object. Alternatively when returnIndex = TRUE, a numeric vector to be used with the original object.

Examples

# Subsetting an object
pbmc <- downsample(pbmc)
# Creating a subsetting index
sampleIdx <- downsample(pbmcPlot, balance = "leiden_cluster",
                        maxCells = 10, returnIndex = TRUE)
plotClusterDimRed(pbmcPlot, cellIdx = sampleIdx)

Description

Export the predicted gene-pair interactions calculated by upstream function linkGenesAndPeaks into an Interact Track file which is compatible with UCSC Genome Browser.

Usage

exportInteractTrack(
  corrMat,
  pathToCoords,
  useGenes = NULL,
  outputPath = getwd()
)

Arguments

Value

No return value. A file located at outputPath will be created.

Examples

bmmc <- normalize(bmmc)
bmmc <- selectGenes(bmmc)
bmmc <- scaleNotCenter(bmmc)
if (requireNamespace("RcppPlanc", quietly = TRUE) &&
    requireNamespace("GenomicRanges", quietly = TRUE) &&
    requireNamespace("IRanges", quietly = TRUE) &&
    requireNamespace("psych", quietly = TRUE)) {
    bmmc <- runINMF(bmmc)
    bmmc <- quantileNorm(bmmc)
    bmmc <- normalizePeak(bmmc)
    bmmc <- imputeKNN(bmmc, reference = "atac", queries = "rna")
    corr <- linkGenesAndPeaks(
        bmmc, useDataset = "rna",
        pathToCoords = system.file("extdata/hg19_genes.bed", package = "rliger")
    )
    resultPath <- tempfile()
    exportInteractTrack(
        corrMat = corr,
        pathToCoords = system.file("extdata/hg19_genes.bed", package = "rliger"),
        outputPath = resultPath
    )
    head(read.table(resultPath, skip = 1))
}

Description

Applies various filters to genes on the shared ($W$) and dataset-specific ($V$) components of the factorization, before selecting those which load most significantly on each factor (in a shared or dataset-specific way).

Usage

getFactorMarkers(
  object,
  dataset1,
  dataset2,
  factorShareThresh = 10,
  datasetSpecificity = NULL,
  logFCThresh = 1,
  pvalThresh = 0.05,
  nGenes = 30,
  printGenes = FALSE,
  verbose = getOption("ligerVerbose", TRUE),
  factor.share.thresh = factorShareThresh,
  dataset.specificity = datasetSpecificity,
  log.fc.thresh = logFCThresh,
  pval.thresh = pvalThresh,
  num.genes = nGenes,
  print.genes = printGenes
)

Arguments

Value

A list object consisting of the following entries:

Examples

library(dplyr)
result <- getFactorMarkers(pbmcPlot, dataset1 = "ctrl", dataset2 = "stim")
print(class(result))
print(names(result))
result$shared %>% group_by(factor_num) %>% top_n(2, logFC)

Description

Calculates proportion of mitochondrial contribution based on raw or normalized data.

Usage

getProportionMito(object, use.norm = FALSE, pattern = "^mt-")

Arguments

Value

Named vector containing proportion of mitochondrial contribution for each cell.

Note

getProportionMito will be deprecated because runGeneralQC generally covers and expands its use case.

Examples

# Example dataset does not contain MT genes, expected to see a message
pbmc$mito <- getProportionMito(pbmc)

Description

h5 calculation wrapper, that runs specified calculation with on-disk matrix in chunks

Usage

H5Apply(
  object,
  FUN,
  init = NULL,
  useData = c("rawData", "normData"),
  chunkSize = 1000,
  verbose = getOption("ligerVerbose"),
  ...
)

Arguments

Details

The FUN function has to have the first four arguments ordered by:

1. chunk data: A sparse matrix (dgCMatrix-class) containing maximum chunkSize columns.

2. x-vector index: The index that subscribes the vector of x slot of a dgCMatrix, which points to the values in each chunk. Mostly used when need to write a new sparse matrix to H5 file.

3. cell index: The column index of each chunk out of the whole original matrix

4. Initialized result: A customized object, the value passed to H5Apply(init) argument will be passed here in the first iteration. And the returned value of FUN will be iteratively passed here in next chunk iterations. So it is important to keep the object structure of the returned value consistent with init.

No default value to these four arguments should be pre-defined because H5Apply will automatically generate the input.

Description

These are functions to download example datasets that are subset from public data.

• PBMC - Downsampled from GSE96583, Kang et al, Nature Biotechnology, 2018. Contains two scRNAseq datasets.

• BMMC - Downsampled from GSE139369, Granja et al, Nature Biotechnology, 2019. Contains two scRNAseq datasets and one scATAC data.

• CGE - Downsampled from GSE97179, Luo et al, Science, 2017. Contains one scRNAseq dataset and one DNA methylation data.

Usage

importPBMC(
  dir = getwd(),
  overwrite = FALSE,
  method = "libcurl",
  verbose = getOption("ligerVerbose", TRUE),
  ...
)

importBMMC(
  dir = getwd(),
  overwrite = FALSE,
  method = "libcurl",
  verbose = getOption("ligerVerbose", TRUE),
  ...
)

importCGE(
  dir = getwd(),
  overwrite = FALSE,
  method = "libcurl",
  verbose = getOption("ligerVerbose", TRUE),
  ...
)

Arguments

Value

Constructed liger object with QC performed and missing data removed.

Examples

pbmc <- importPBMC()
bmmc <- importBMMC()
cge <- importCGE()

Description

This function is designed for creating peak data for a dataset with only gene expression. This function uses quantile normalized cell factor loading to find nearest neighbors between cells from the queried dataset (without peak) and cells from reference dataset (with peak). And then impute the peak for the former basing on the weight. Therefore, the reference dataset selected must be of "atac" modality setting.

Usage

imputeKNN(
  object,
  reference,
  queries = NULL,
  nNeighbors = 20,
  weight = TRUE,
  norm = TRUE,
  scale = FALSE,
  verbose = getOption("ligerVerbose", TRUE),
  ...,
  knn_k = nNeighbors
)

Arguments

Value

The input object where queried ligerDataset objects in datasets slot are replaced. These datasets will all be converted to ligerATACDataset class with an additional slot rawPeak to store the imputed peak counts, and normPeak for normalized imputed peak counts if norm = TRUE.

Examples

bmmc <- normalize(bmmc)
bmmc <- selectGenes(bmmc, datasets.use = "rna")
bmmc <- scaleNotCenter(bmmc)
if (requireNamespace("RcppPlanc", quietly = TRUE)) {
    bmmc <- runINMF(bmmc, k = 20)
    bmmc <- quantileNorm(bmmc)
    bmmc <- normalizePeak(bmmc)
    bmmc <- imputeKNN(bmmc, reference = "atac", queries = "rna")
}

Description

Check if given liger object if under new implementation

Usage

is.newLiger(object)

Arguments

Value

TRUE if the version of object is later than or equal to 1.99.0. Otherwise FALSE. It raises an error if input object is not of liger class.

Examples

is.newLiger(pbmc) # TRUE

Description

Check if a liger or ligerDataset object is made of HDF5 file

Usage

isH5Liger(object, dataset = NULL)

Arguments

Value

TRUE or FALSE for the specified check.

Examples

isH5Liger(pbmc)
isH5Liger(pbmc, "ctrl")
ctrl <- dataset(pbmc, "ctrl")
isH5Liger(ctrl)

Description

liger object is the main data container for LIGER analysis in R. The slot datasets is a list where each element should be a ligerDataset object containing dataset specific information, such as the expression matrices. The other parts of liger object stores information that can be shared across the analysis, such as the cell metadata and factorization result matrices.

This manual provides explanation to the liger object structure as well as usage of class-specific methods. Please see detail sections for more information.

For liger objects created with older versions of rliger package, please try updating the objects individually with convertOldLiger.

Usage

datasets(x, check = NULL)

datasets(x, check = TRUE) <- value

dataset(x, dataset = NULL)

dataset(x, dataset, type = NULL, qc = TRUE) <- value

cellMeta(
  x,
  columns = NULL,
  useDatasets = NULL,
  cellIdx = NULL,
  as.data.frame = FALSE,
  ...
)

cellMeta(
  x,
  columns = NULL,
  useDatasets = NULL,
  cellIdx = NULL,
  inplace = FALSE,
  check = FALSE
) <- value

defaultCluster(x, useDatasets = NULL, ...)

defaultCluster(x, name = NULL, useDatasets = NULL, ...) <- value

dimReds(x)

dimReds(x) <- value

dimRed(x, name = NULL, useDatasets = NULL, cellIdx = NULL, ...)

dimRed(x, name = NULL, useDatasets = NULL, cellIdx = NULL, ...) <- value

defaultDimRed(x, useDatasets = NULL, cellIdx = NULL)

defaultDimRed(x) <- value

varFeatures(x)

varFeatures(x, check = TRUE) <- value

varUnsharedFeatures(x, dataset = NULL)

varUnsharedFeatures(x, dataset, check = TRUE) <- value

commands(x, funcName = NULL, arg = NULL)

## S4 method for signature 'liger'
show(object)

## S4 method for signature 'liger'
dim(x)

## S4 method for signature 'liger'
dimnames(x)

## S4 replacement method for signature 'liger,list'
dimnames(x) <- value

## S4 method for signature 'liger'
datasets(x, check = NULL)

## S4 replacement method for signature 'liger,logical'
datasets(x, check = TRUE) <- value

## S4 replacement method for signature 'liger,missing'
datasets(x, check = TRUE) <- value

## S4 method for signature 'liger,character_OR_NULL'
dataset(x, dataset = NULL)

## S4 method for signature 'liger,missing'
dataset(x, dataset = NULL)

## S4 method for signature 'liger,numeric'
dataset(x, dataset = NULL)

## S4 replacement method for signature 'liger,character,missing,ANY,ligerDataset'
dataset(x, dataset, type = NULL, qc = TRUE) <- value

## S4 replacement method for signature 'liger,character,ANY,ANY,matrixLike'
dataset(x, dataset, type = c("rawData", "normData"), qc = FALSE) <- value

## S4 replacement method for signature 'liger,character,missing,ANY,NULL'
dataset(x, dataset, type = NULL, qc = TRUE) <- value

## S3 method for class 'liger'
names(x)

## S3 replacement method for class 'liger'
names(x) <- value

## S3 method for class 'liger'
length(x)

## S3 method for class 'liger'
lengths(x, use.names = TRUE)

## S4 method for signature 'liger,NULL'
cellMeta(
  x,
  columns = NULL,
  useDatasets = NULL,
  cellIdx = NULL,
  as.data.frame = FALSE,
  ...
)

## S4 method for signature 'liger,character'
cellMeta(
  x,
  columns = NULL,
  useDatasets = NULL,
  cellIdx = NULL,
  as.data.frame = FALSE,
  ...
)

## S4 method for signature 'liger,missing'
cellMeta(
  x,
  columns = NULL,
  useDatasets = NULL,
  cellIdx = NULL,
  as.data.frame = FALSE,
  ...
)

## S4 replacement method for signature 'liger,missing'
cellMeta(x, columns = NULL, useDatasets = NULL, cellIdx = NULL, check = FALSE) <- value

## S4 replacement method for signature 'liger,character'
cellMeta(
  x,
  columns = NULL,
  useDatasets = NULL,
  cellIdx = NULL,
  inplace = TRUE,
  check = FALSE
) <- value

## S4 method for signature 'liger'
rawData(x, dataset = NULL)

## S4 replacement method for signature 'liger,ANY,ANY,matrixLike_OR_NULL'
rawData(x, dataset = NULL, check = TRUE) <- value

## S4 replacement method for signature 'liger,ANY,ANY,H5D'
rawData(x, dataset = NULL, check = TRUE) <- value

## S4 method for signature 'liger'
normData(x, dataset = NULL)

## S4 replacement method for signature 'liger,ANY,ANY,matrixLike_OR_NULL'
normData(x, dataset = NULL, check = TRUE) <- value

## S4 replacement method for signature 'liger,ANY,ANY,H5D'
normData(x, dataset = NULL, check = TRUE) <- value

## S4 method for signature 'liger,ANY'
scaleData(x, dataset = NULL)

## S4 replacement method for signature 'liger,ANY,ANY,matrixLike_OR_NULL'
scaleData(x, dataset = NULL, check = TRUE) <- value

## S4 replacement method for signature 'liger,ANY,ANY,H5D'
scaleData(x, dataset = NULL, check = TRUE) <- value

## S4 replacement method for signature 'liger,ANY,ANY,H5Group'
scaleData(x, dataset = NULL, check = TRUE) <- value

## S4 method for signature 'liger,character'
scaleUnsharedData(x, dataset = NULL)

## S4 method for signature 'liger,numeric'
scaleUnsharedData(x, dataset = NULL)

## S4 replacement method for signature 'liger,ANY,ANY,matrixLike_OR_NULL'
scaleUnsharedData(x, dataset = NULL, check = TRUE) <- value

## S4 replacement method for signature 'liger,ANY,ANY,H5D'
scaleUnsharedData(x, dataset = NULL, check = TRUE) <- value

## S4 replacement method for signature 'liger,ANY,ANY,H5Group'
scaleUnsharedData(x, dataset = NULL, check = TRUE) <- value

## S4 method for signature 'liger,ANY,ANY,ANY'
getMatrix(
  x,
  slot = c("rawData", "normData", "scaleData", "scaleUnsharedData", "H", "V", "U", "A",
    "B", "W", "H.norm"),
  dataset = NULL,
  returnList = FALSE
)

## S4 method for signature 'liger,ANY'
getH5File(x, dataset = NULL)

## S3 replacement method for class 'liger'
x[[i]] <- value

## S3 method for class 'liger'
x$name

## S3 replacement method for class 'liger'
x$name <- value

## S4 method for signature 'liger'
defaultCluster(x, useDatasets = NULL, droplevels = FALSE, ...)

## S4 replacement method for signature 'liger,ANY,ANY,character'
defaultCluster(x, name = NULL, useDatasets = NULL, ...) <- value

## S4 replacement method for signature 'liger,ANY,ANY,factor'
defaultCluster(x, name = NULL, useDatasets = NULL, droplevels = TRUE, ...) <- value

## S4 replacement method for signature 'liger,ANY,ANY,NULL'
defaultCluster(x, name = NULL, useDatasets = NULL, ...) <- value

## S4 method for signature 'liger'
dimReds(x)

## S4 replacement method for signature 'liger,list'
dimReds(x) <- value

## S4 method for signature 'liger,missing_OR_NULL'
dimRed(x, name = NULL, useDatasets = NULL, cellIdx = NULL, ...)

## S4 method for signature 'liger,index'
dimRed(x, name = NULL, useDatasets = NULL, cellIdx = NULL, ...)

## S4 replacement method for signature 'liger,index,ANY,ANY,NULL'
dimRed(x, name = NULL, useDatasets = NULL, cellIdx = NULL, ...) <- value

## S4 replacement method for signature 'liger,character,ANY,ANY,matrixLike'
dimRed(
  x,
  name = NULL,
  useDatasets = NULL,
  cellIdx = NULL,
  asDefault = NULL,
  inplace = FALSE,
  ...
) <- value

## S4 method for signature 'liger'
defaultDimRed(x, useDatasets = NULL, cellIdx = cellIdx)

## S4 replacement method for signature 'liger,character'
defaultDimRed(x) <- value

## S4 method for signature 'liger'
varFeatures(x)

## S4 replacement method for signature 'liger,ANY,character'
varFeatures(x, check = TRUE) <- value

## S4 method for signature 'liger,ANY'
varUnsharedFeatures(x, dataset = NULL)

## S4 replacement method for signature 'liger,ANY,ANY,character'
varUnsharedFeatures(x, dataset, check = TRUE) <- value

## S3 method for class 'liger'
fortify(model, data, ...)

## S3 method for class 'liger'
c(...)

## S4 method for signature 'liger'
commands(x, funcName = NULL, arg = NULL)

## S4 method for signature 'ligerDataset,missing'
varUnsharedFeatures(x, dataset = NULL)

## S4 replacement method for signature 'ligerDataset,missing,ANY,character'
varUnsharedFeatures(x, dataset = NULL, check = TRUE) <- value

Arguments

Value

See detailed sections for explanetion.

Input liger object updated with replaced/new variable in cellMeta(x).

Slots

datasets

list of ligerDataset objects. Use generic dataset, dataset<-, datasets or datasets<- to interact with. See detailed section accordingly.

cellMeta

DFrame object for cell metadata. Pre-existing metadata, QC metrics, cluster labeling, low-dimensional embedding and etc. are all stored here. Use generic cellMeta, cellMeta<-, $, [[]] or [[]]<- to interact with. See detailed section accordingly.

varFeatures

Character vector of feature names. Use generic varFeatures or varFeatures<- to interact with. See detailed section accordingly.

W

Matrix of gene loading for each factor. See runIntegration.

H.norm

Matrix of aligned factor loading for each cell. See quantileNorm and runIntegration.

commands

List of ligerCommand objects. Record of analysis. Use commands to retrieve information. See detailed section accordingly.

uns

List for unstructured meta-info of analyses or presets.

version

Record of version of rliger package

Dataset access

datasets() method only accesses the datasets slot, the list of ligerDataset objects. dataset() method accesses a single dataset, with subsequent cell metadata updates and checks bonded when adding or modifying a dataset. Therefore, when users want to modify something inside a ligerDataset while no cell metadata change should happen, it is recommended to use: datasets(x)[[name]] <- ligerD for efficiency, though the result would be the same as dataset(x, name) <- ligerD.

length() and names() methods are implemented to access the number and names of datasets. names<- method is supported for modifying dataset names, with taking care of the "dataset" variable in cell metadata.

Matrix access

For liger object, rawData(), normData, scaleData() and scaleUnsharedData() methods are exported for users to access the corresponding feature expression matrix with specification of one dataset. For retrieving a type of matrix from multiple datasets, please use getMatrix() method.

When only one matrix is expected to be retrieved by getMatrix(), the matrix itself will be returned. A list will be returned if multiple matrices is requested (by querying multiple datasets) or returnList is set to TRUE.

Cell metadata access

Three approaches are provided for access of cell metadata. A generic function cellMeta is implemented with plenty of options and multi-variable accessibility. Besides, users can use double-bracket (e.g. ligerObj[[varName]]) or dollor-sign (e.g. ligerObj$nUMI) to access or modify single variables.

For users' convenience of generating a customized ggplot with available cell metadata, the S3 method fortify.liger is implemented. With this under the hook, users can create simple ggplots by directly starting with ggplot(ligerObj, aes(...)) where cell metadata variables can be directly thrown into aes().

Special partial metadata insertion is implemented specifically for mapping categorical annotation from sub-population (subset object) back to original experiment (full-size object). For example, when sub-clustering and annotation is done for a specific cell-type of cells (stored in subobj) subset from an experiment (stored as obj), users can do cellMeta(obj, "sub_ann", cellIdx = colnames(subobj)) <- subobj$sub_ann to map the value back, leaving other cells non-annotated with NAs. Plotting with this variable will then also show NA cells with default grey color. Furthermore, sub-clustering labels for other cell types can also be mapped to the same variable. For example, cellMeta(obj, "sub_ann", cellIdx = colnames(subobj2)) <- subobj2$sub_ann. As long as the labeling variables are stored as factor class (categorical), the levels (category names) will be properly handled and merged. Other situations follow the R default behavior (e.g. categories might be converted to integer numbers if mapped to numerical variable in the original object). Note that this feature is only available with using the generic function cellMeta but not with the `[[` or `$` accessing methods due to syntax reasons.

The generic defaultCluster works as both getter and setter. As a setter, users can do defaultCluster(obj) <- "existingVariableName" to set a categorical variable as default cluster used for visualization or downstream analysis. Users can also do defaultCluster(obj, "newVarName") <- factorOfLabels to push new labeling into the object and set as default. For getter method, the function returns a factor object of the default cluster labeling. Argument useDatasets can be used for requiring that given or retrieved labeling should match with cells in specified datasets. We generally don't recommend setting "dataset" as a default cluster because it is a preserved (always existing) field in metadata and can lead to meaningless result when running analysis that utilizes both clustering information and the dataset source information.

Dimension reduction access

Currently, low-dimensional representaion of cells, presented as dense matrices, are all stored in dimReds slot, and can totally be accessed with generics dimRed and dimRed<-. Adding a dimRed to the object looks as simple as dimRed(obj, "name") <- matrixLike. It can be retrieved back with dimRed(obj, "name"). Similar to having a default cluster labeling, we also constructed the feature of default dimRed. It can be set with defaultDimRed(obj) <- "existingMatLikeVar" and the matrix can be retrieved with defaultDimRed(obj).

Variable feature access

The varFeatures slot allows for character vectors of gene names. varFeatures(x) returns this vector and value for varFeatures<- method has to be a character vector or NULL. The replacement method, when check = TRUE performs checks on gene name consistency check across the scaleData, H, V slots of inner ligerDataset objects as well as the W and H.norm slots of the input liger object.

Command records

rliger functions, that perform calculation and update the liger object, will be recorded in a ligerCommand object and stored in the commands slot, a list, of liger object. Method commands() is implemented to retrieve or show the log history. Running with funcName = NULL (default) returns all command labels. Specifying funcName allows partial matching to all command labels and returns a subset list (of ligerCommand object) of matches (or the ligerCommand object if only one match found). If arg is further specified, a subset list of parameters from the matches will be returned. For example, requesting a list of resolution values used in all louvain cluster attempts: commands(ligerObj, "louvainCluster", "resolution")

Dimensionality

For a liger object, the column orientation is assigned for cells. Due to the data structure, it is hard to define a row index for the liger object, which might contain datasets that vary in number of genes.

Therefore, for liger objects, dim and dimnames returns NA/NULL for rows and total cell counts/barcodes for the columns.

For direct call of dimnames<- method, value should be a list with NULL as the first element and valid cell identifiers as the second element. For colnames<- method, the character vector of cell identifiers. rownames<- method is not applicable.

Subsetting

For more detail of subsetting a liger object or a ligerDataset object, please check out subsetLiger and subsetLigerDataset. Here, we set the S4 method "single-bracket" [ as a quick wrapper to subset a liger object. Note that j serves as cell subscriptor which can be any valid index refering the collection of all cells (i.e. rownames(cellMeta(obj))). While i, the feature subscriptor can only be character vector because the features for each dataset can vary. ... arugments are passed to subsetLiger so that advanced options are allowed.

Combining multiple liger object

The list of datasets slot, the rows of cellMeta slot and the list of commands slot will be simply concatenated. Variable features in varFeatures slot will be taken a union. The $W$ and $H.norm$ matrices are not taken into account for now.

Examples

# Methods for base generics
pbmcPlot
print(pbmcPlot)
dim(pbmcPlot)
ncol(pbmcPlot)
colnames(pbmcPlot)[1:5]
pbmcPlot[varFeatures(pbmcPlot)[1:10], 1:10]
names(pbmcPlot)
length(pbmcPlot)

# rliger generics
## Retrieving dataset(s), replacement methods available
datasets(pbmcPlot)
dataset(pbmcPlot, "ctrl")
dataset(pbmcPlot, 2)

## Retrieving cell metadata, replacement methods available
cellMeta(pbmcPlot)
head(pbmcPlot[["nUMI"]])

## Retrieving dimemtion reduction matrix
head(dimRed(pbmcPlot, "UMAP"))

## Retrieving variable features, replacement methods available
varFeatures(pbmcPlot)

## Command record/history
pbmcPlot <- scaleNotCenter(pbmcPlot)
commands(pbmcPlot)
commands(pbmcPlot, funcName = "scaleNotCenter")

# S3 methods
pbmcPlot2 <- pbmcPlot
names(pbmcPlot2) <- paste0(names(pbmcPlot), 2)
c(pbmcPlot, pbmcPlot2)

library(ggplot2)
ggplot(pbmcPlot, aes(x = UMAP_1, y = UMAP_2)) + geom_point()
cellMeta(pbmc)
# Add new variable
pbmc[["newVar"]] <- 1
cellMeta(pbmc)
# Change existing variable
pbmc[["newVar"]][1:3] <- 1:3
cellMeta(pbmc)

Description

Inherits from ligerDataset class. Contained slots can be referred with the link.

Slots

rawPeak

sparse matrix

normPeak

sparse matrix

Description

ligerCommand object: Record the input and time of a LIGER function call

Usage

## S4 method for signature 'ligerCommand'
show(object)

Arguments

Slots

funcName

Name of the function

time

A time stamp object

call

A character string converted from system call

parameters

List of all arguments except the liger object. Large object are summarized to short string.

objSummary

List of attributes of the liger object as a snapshot when command is operated.

ligerVersion

Character string converted from packageVersion("rliger").

dependencyVersion

Named character vector of version number, if any dependency library has a chance to be included by the function. A dependency might only be invoked under certain conditions, such as using an alternative algorithm, which a call does not actually reach to, but it would still be included for this call.

Examples

pbmc <- normalize(pbmc)
cmd <- commands(pbmc, "normalize")
cmd

Description

Object for storing dastaset specific information. Will be embedded within a higher level liger object

Usage

rawData(x, dataset = NULL)

rawData(x, dataset = NULL, check = TRUE) <- value

normData(x, dataset = NULL)

normData(x, dataset = NULL, check = TRUE) <- value

scaleData(x, dataset = NULL)

scaleData(x, dataset = NULL, check = TRUE) <- value

scaleUnsharedData(x, dataset = NULL)

scaleUnsharedData(x, dataset = NULL, check = TRUE) <- value

getMatrix(x, slot = "rawData", dataset = NULL, returnList = FALSE)

h5fileInfo(x, info = NULL)

h5fileInfo(x, info = NULL, check = TRUE) <- value

getH5File(x, dataset = NULL)

## S4 method for signature 'ligerDataset,missing'
getH5File(x, dataset = NULL)

featureMeta(x, check = NULL)

featureMeta(x, check = TRUE) <- value

## S4 method for signature 'ligerDataset'
show(object)

## S4 method for signature 'ligerDataset'
dim(x)

## S4 method for signature 'ligerDataset'
dimnames(x)

## S4 replacement method for signature 'ligerDataset,list'
dimnames(x) <- value

## S4 method for signature 'ligerDataset'
rawData(x, dataset = NULL)

## S4 replacement method for signature 'ligerDataset,ANY,ANY,matrixLike_OR_NULL'
rawData(x, dataset = NULL, check = TRUE) <- value

## S4 replacement method for signature 'ligerDataset,ANY,ANY,H5D'
rawData(x, dataset = NULL, check = TRUE) <- value

## S4 method for signature 'ligerDataset'
normData(x, dataset = NULL)

## S4 replacement method for signature 'ligerDataset,ANY,ANY,matrixLike_OR_NULL'
normData(x, dataset = NULL, check = TRUE) <- value

## S4 replacement method for signature 'ligerDataset,ANY,ANY,H5D'
normData(x, dataset = NULL, check = TRUE) <- value

## S4 method for signature 'ligerDataset,missing'
scaleData(x, dataset = NULL)

## S4 replacement method for signature 'ligerDataset,ANY,ANY,matrixLike_OR_NULL'
scaleData(x, dataset = NULL, check = TRUE) <- value

## S4 replacement method for signature 'ligerDataset,ANY,ANY,H5D'
scaleData(x, dataset = NULL, check = TRUE) <- value

## S4 replacement method for signature 'ligerDataset,ANY,ANY,H5Group'
scaleData(x, dataset = NULL, check = TRUE) <- value

## S4 method for signature 'ligerDataset,missing'
scaleUnsharedData(x, dataset = NULL)

## S4 replacement method for signature 'ligerDataset,missing,ANY,matrixLike_OR_NULL'
scaleUnsharedData(x, check = TRUE) <- value

## S4 replacement method for signature 'ligerDataset,missing,ANY,H5D'
scaleUnsharedData(x, check = TRUE) <- value

## S4 replacement method for signature 'ligerDataset,missing,ANY,H5Group'
scaleUnsharedData(x, check = TRUE) <- value

## S4 method for signature 'ligerDataset,ANY,missing,missing'
getMatrix(
  x,
  slot = c("rawData", "normData", "scaleData", "scaleUnsharedData", "H", "V", "U", "A",
    "B"),
  dataset = NULL
)

## S4 method for signature 'ligerDataset'
h5fileInfo(x, info = NULL)

## S4 replacement method for signature 'ligerDataset'
h5fileInfo(x, info = NULL, check = TRUE) <- value

## S4 method for signature 'ligerDataset'
featureMeta(x, check = NULL)

## S4 replacement method for signature 'ligerDataset'
featureMeta(x, check = TRUE) <- value

## S3 method for class 'ligerDataset'
cbind(x, ..., deparse.level = 1)

Arguments

Slots

rawData

Raw data.

normData

Normalized data

scaleData

Scaled data, usually with subset variable features

scaleUnsharedData

Scaled data of features not shared with other datasets

varUnsharedFeatures

Variable features not shared with other datasets

V

matrix

A

matrix

B

matrix

H

matrix

U

matrix

h5fileInfo

list

featureMeta

Feature metadata, DataFrame

colnames

character

rownames

character

Matrix access

For ligerDataset object, rawData(), normData, scaleData() and scaleUnsharedData() methods are exported for users to access the corresponding feature expression matrix. Replacement methods are also available to modify the slots.

For other matrices, such as the $H$ and $V$, which are dataset specific, please use getMatrix() method with specifying slot name. Directly accessing slot with @ is generally not recommended.

H5 file and information access

A ligerDataset object has a slot called h5fileInfo, which is a list object. The first element is called $H5File, which is an H5File class object and is the connection to the input file. The second element is $filename which stores the absolute path of the H5 file in the current machine. The third element $formatType stores the name of preset being used, if applicable. The other following keys pair with paths in the H5 file that point to specific data for constructing a feature expression matrix.

h5fileInfo() method access the list described above and simply retrieves the corresponding value. When info = NULL, returns the whole list. When length(info) == 1, returns the requested list value. When more info requested, returns a subset list.

The replacement method modifies the list elements and corresponding slot value (if applicable) at the same time. For example, running h5fileInfo(obj, "rawData") <- newPath not only updates the list, but also updates the rawData slot with the H5D class data at "newPath" in the H5File object.

getH5File() is a wrapper and is equivalent to h5fileInfo(obj, "H5File").

Feature metadata access

A slot featureMeta is included for each ligerDataset object. This slot requires a DataFrame-class object, which is the same as cellMeta slot of a liger object. However, the associated S4 methods only include access to the whole table for now. Internal information access follows the same way as data.frame operation. For example, featureMeta(ligerD)$nCell or featureMeta(ligerD)[varFeatures(ligerObj), "gene_var"].

Dimensionality

For a ligerDataset object, the column orientation is assigned for cells and rows are for features. Therefore, for ligerDataset objects, dim() returns a numeric vector of two numbers which are number of features and number of cells. dimnames() returns a list of two character vectors, which are the feature names and the cell barcodes.

For direct call of dimnames<- method, value should be a list with a character vector of feature names as the first element and cell identifiers as the second element. For colnames<- method, the character vector of cell identifiers. For rownames<- method, the character vector of feature names.

Subsetting

For more detail of subsetting a liger object or a ligerDataset object, please check out subsetLiger and subsetLigerDataset. Here, we set the S3 method "single-bracket" [ as a quick wrapper to subset a ligerDataset object. i and j serves as feature and cell subscriptor, respectively, which can be any valid index refering the available features and cells in a dataset. ... arugments are passed to subsetLigerDataset so that advanced options are allowed.

Concatenate ligerDataset

cbind() method is implemented for concatenating ligerDataset objects by cells. When applying, all feature expression matrix will be merged with taking a union of all features for the rows.

Examples

ctrl <- dataset(pbmc, "ctrl")

# Methods for base generics
ctrl
print(ctrl)
dim(ctrl)
ncol(ctrl)
nrow(ctrl)
colnames(ctrl)[1:5]
rownames(ctrl)[1:5]
ctrl[1:5, 1:5]

# rliger generics
## raw data
m <- rawData(ctrl)
class(m)
dim(m)
## normalized data
pbmc <- normalize(pbmc)
ctrl <- dataset(pbmc, "ctrl")
m <- normData(ctrl)
class(m)
dim(m)
## scaled data
pbmc <- selectGenes(pbmc)
pbmc <- scaleNotCenter(pbmc)
ctrl <- dataset(pbmc, "ctrl")
m <- scaleData(ctrl)
class(m)
dim(m)
n <- scaleData(pbmc, "ctrl")
identical(m, n)
## Any other matrices
if (requireNamespace("RcppPlanc", quietly = TRUE)) {
    pbmc <- runOnlineINMF(pbmc, k = 20, minibatchSize = 100)
    ctrl <- dataset(pbmc, "ctrl")
    V <- getMatrix(ctrl, "V")
    V[1:5, 1:5]
    Vs <- getMatrix(pbmc, "V")
    length(Vs)
    names(Vs)
    identical(Vs$ctrl, V)
}

Description

Inherits from ligerDataset class. Contained slots can be referred with the link. scaleNotCenter applied on datasets of this class will automatically be taken by reversing the normalized data instead of scaling the variable features.

Description

Inherits from ligerDataset class. Contained slots can be referred with the link. This subclass does not have any different from the default ligerDataset class except the class name.

Description

Inherits from ligerDataset class. Contained slots can be referred with the link.

Slots

coordinate

dense matrix

Description

For converting a liger object to a Seurat object, the rawData, normData, and scaleData from each dataset, the cellMeta, H.norm and varFeatures slot will be included. Compatible with V4 and V5. It is not recommended to use this conversion if your liger object contains datasets from various modalities.

Usage

ligerToSeurat(
  object,
  assay = NULL,
  identByDataset = FALSE,
  merge = FALSE,
  nms = NULL,
  renormalize = NULL,
  use.liger.genes = NULL,
  by.dataset = identByDataset
)

Arguments

Value

Always returns Seurat object(s) of the latest version. By default a Seurat object with split layers, e.g. with layers like "counts.ctrl" and "counts.stim". If merge = TRUE, return a single Seurat object with layers for all datasets merged.

Examples

seu <- ligerToSeurat(pbmc)

Description

Evaluate the relationships between pairs of genes and peaks based on specified distance metric. Usually used for inferring the correlation between gene expression and imputed peak counts for datasets without the modality originally (i.e. applied to imputeKNN result).

Usage

linkGenesAndPeaks(
  object,
  useDataset,
  pathToCoords,
  useGenes = NULL,
  method = c("spearman", "pearson", "kendall"),
  alpha = 0.05,
  verbose = getOption("ligerVerbose", TRUE),
  path_to_coords = pathToCoords,
  genes.list = useGenes,
  dist = method
)

Arguments

Value

A sparse matrix with peak names as rows and gene names as columns, with each element indicating the correlation between peak i and gene j, 0 if the gene and peak are not significantly linked.

See Also

imputeKNN

Examples

if (requireNamespace("RcppPlanc", quietly = TRUE) &&
    requireNamespace("GenomicRanges", quietly = TRUE) &&
    requireNamespace("IRanges", quietly = TRUE) &&
    requireNamespace("psych", quietly = TRUE)) {
    bmmc <- normalize(bmmc)
    bmmc <- selectGenes(bmmc)
    bmmc <- scaleNotCenter(bmmc)
    bmmc <- runINMF(bmmc, miniBatchSize = 100)
    bmmc <- quantileNorm(bmmc)
    bmmc <- normalizePeak(bmmc)
    bmmc <- imputeKNN(bmmc, reference = "atac", queries = "rna")
    corr <- linkGenesAndPeaks(
        bmmc, useDataset = "rna",
        pathToCoords = system.file("extdata/hg19_genes.bed", package = "rliger")
    )
}

Description

After quantile normalization, users can additionally run the Louvain algorithm for community detection, which is widely used in single-cell analysis and excels at merging small clusters into broad cell classes.

Arguments

Value

object with refined cluster assignment updated in "louvain_cluster" variable in cellMeta slot. Can be fetched with object$louvain_cluster

See Also

rliger-deprecated

Description

Fast calculation of feature count matrix

Usage

makeFeatureMatrix(bedmat, barcodes)

Arguments

Value

A feature count matrix with features as rows and barcodes as columns

Examples

## Not run: 
gene.counts <- makeFeatureMatrix(genes.bc, barcodes)
promoter.counts <- makeFeatureMatrix(promoters.bc, barcodes)
samnple <- gene.counts + promoter.counts

## End(Not run)

Description

Export the predicted gene-pair interactions calculated by upstream function linkGenesAndPeaks into an Interact Track file which is compatible with UCSC Genome Browser.

Arguments

Value

No return value. A file located at outputPath will be created.

See Also

rliger-deprecated, exportInteractTrack

Description

Creates a riverplot to show how separate cluster assignments from two datasets map onto a joint clustering. The joint clustering is by default the object clustering, but an external one can also be passed in. Uses the riverplot package to construct riverplot object and then plot.

Arguments

Value

object with refined cluster assignment updated in "louvain_cluster" variable in cellMeta slot. Can be fetched with object$louvain_cluster

See Also

rliger-deprecated

Description

Designed for fast variable creation when a new variable is going to be created from existing variable. For example, multiple samples can be mapped to the same study design condition, clusters can be mapped to cell types.

Usage

mapCellMeta(object, from, newTo = NULL, ...)

Arguments

Value

When newTo = NULL, a factor object of the new variable. Otherwise, the input object with variable newTo updated in cellMeta(object).

Examples

pbmc <- mapCellMeta(pbmc, from = "dataset", newTo = "modal",
                    ctrl = "rna", stim = "rna")

Description

This function merges hdf5 files generated from different libraries (cell ranger by default) before they are preprocessed through Liger pipeline.

Usage

mergeH5(
  file.list,
  library.names,
  new.filename,
  format.type = "10X",
  data.name = NULL,
  indices.name = NULL,
  indptr.name = NULL,
  genes.name = NULL,
  barcodes.name = NULL
)

Arguments

Value

Directly generates newly merged hdf5 file.

Examples

## Not run: 
# For instance, we want to merge two datasets saved in HDF5 files (10X
# CellRanger) paths to datasets: "library1.h5","library2.h5"
# dataset names: "lib1", "lib2"
# name for output HDF5 file: "merged.h5"
mergeH5(list("library1.h5","library2.h5"), c("lib1","lib2"), "merged.h5")

## End(Not run)

Description

mergeSparseAll takes in a list of DGEs, with genes as rows and cells as columns, and merges them into a single DGE. Also adds libraryNames to colnames from each DGE if expected to be overlap (common with 10X barcodes). Values in rawData or normData slot of a ligerDataset object can be processed with this.

For a list of dense matrices, usually the values in scaleData slot of a ligerDataset object, please use mergeDenseAll which works in the same way.

Usage

mergeSparseAll(
  datalist,
  libraryNames = NULL,
  mode = c("union", "intersection")
)

mergeDenseAll(datalist, libraryNames = NULL)

Arguments

Value

dgCMatrix or matrix with all barcodes in datalist as columns and the union of genes in datalist as rows.

Examples

rawDataList <- getMatrix(pbmc, "rawData")
merged <- mergeSparseAll(rawDataList, libraryNames = names(pbmc))

Description

Return preset modality of a ligerDataset object or that of all datasets in a liger object

Usage

modalOf(object)

Arguments

Value

A single character of modality setting value for ligerDataset object, or a named vector for liger object, where the names are dataset names.

Examples

modalOf(pbmc)
ctrl <- dataset(pbmc, "ctrl")
modalOf(ctrl)
ctrl.atac <- as.ligerDataset(ctrl, modal = "atac")
modalOf(ctrl.atac)

Description

Perform library size normalization on raw counts input. As for the preprocessing step of iNMF integration, by default we don't multiply the normalized values with a scale factor, nor do we take the log transformation. Applicable S3 methods can be found in Usage section.

normalizePeak is designed for datasets of "atac" modality, i.e. stored in ligerATACDataset. S3 method for various container object is not supported yet due to difference in architecture design.

Usage

normalize(object, ...)

## S3 method for class 'dgCMatrix'
normalize(object, log = FALSE, scaleFactor = NULL, ...)

## S3 method for class 'ligerDataset'
normalize(object, chunk = 1000, verbose = getOption("ligerVerbose", TRUE), ...)

## S3 method for class 'liger'
normalize(
  object,
  useDatasets = NULL,
  verbose = getOption("ligerVerbose", TRUE),
  format.type = NULL,
  remove.missing = NULL,
  ...
)

## S3 method for class 'Seurat'
normalize(object, assay = NULL, layer = "counts", save = "ligerNormData", ...)

normalizePeak(
  object,
  useDatasets = NULL,
  verbose = getOption("ligerVerbose", TRUE),
  ...
)

Arguments

Value

Updated object.

• dgCMatrix method - Returns processed dgCMatrix object

• ligerDataset method - Updates the normData slot of the object

• liger method - Updates the normData slot of chosen datasets

• Seurat method - Adds a named layer in chosen assay (V5), or update the data slot of the chosen assay (<=V4)

• normalizePeak - Updates the normPeak slot of chosen datasets.

Examples

pbmc <- normalize(pbmc)

Description

Please turn to runOnlineINMF or runIntegration.

Perform online integrative non-negative matrix factorization to represent multiple single-cell datasets in terms of H, W, and V matrices. It optimizes the iNMF objective function using online learning (non-negative least squares for H matrix, hierarchical alternating least squares for W and V matrices), where the number of factors is set by k. The function allows online learning in 3 scenarios: (1) fully observed datasets; (2) iterative refinement using continually arriving datasets; and (3) projection of new datasets without updating the existing factorization. All three scenarios require fixed memory independent of the number of cells.

For each dataset, this factorization produces an H matrix (cells by k), a V matrix (k by genes), and a shared W matrix (k by genes). The H matrices represent the cell factor loadings. W is identical among all datasets, as it represents the shared components of the metagenes across datasets. The V matrices represent the dataset-specific components of the metagenes.

Arguments

Value

liger object with H, W, V, A and B slots set.

Description

Please turn to runINMF or runIntegration.

Perform integrative non-negative matrix factorization to return factorized H, W, and V matrices. It optimizes the iNMF objective function using block coordinate descent (alternating non-negative least squares), where the number of factors is set by k. TODO: include objective function equation here in documentation (using deqn)

For each dataset, this factorization produces an H matrix (cells by k), a V matrix (k by genes), and a shared W matrix (k by genes). The H matrices represent the cell factor loadings. W is held consistent among all datasets, as it represents the shared components of the metagenes across datasets. The V matrices represent the dataset-specific components of the metagenes.

Arguments

Value

liger object with H, W, and V slots set.

See Also

rliger-deprecated

Description

Uses an efficient strategy for updating that takes advantage of the information in the existing factorization. Assumes that variable features are presented in the new datasets. Two modes are supported (controlled by merge):

• Append new data to existing datasets specified by useDatasets. Here the existing $V$ matrices for the target datasets will directly be used as initialization, and new $H$ matrices for the merged matrices will be initialized accordingly.

• Set new data as new datasets. Initial $V$ matrices for them will be copied from datasets specified by useDatasets, and new $H$ matrices will be initialized accordingly.

Usage

optimizeNewData(
  object,
  dataNew,
  useDatasets,
  merge = TRUE,
  lambda = NULL,
  nIteration = 30,
  seed = 1,
  verbose = getOption("ligerVerbose"),
  new.data = dataNew,
  which.datasets = useDatasets,
  add.to.existing = merge,
  max.iters = nIteration,
  thresh = NULL
)

Arguments

Value

object with W slot updated with the new $W$ matrix, and the H and V slots of each ligerDataset object in the datasets slot updated with the new dataset specific $H$ and $V$ matrix, respectively.

See Also

runINMF, optimizeNewK, optimizeNewLambda

Examples

pbmc <- normalize(pbmc)
pbmc <- selectGenes(pbmc)
pbmc <- scaleNotCenter(pbmc)
# Only running a few iterations for fast examples
if (requireNamespace("RcppPlanc", quietly = TRUE)) {
    pbmc <- runINMF(pbmc, k = 20, nIteration = 2)
    # Create fake new data by increasing all non-zero count in "ctrl" by 1,
    # and make unique cell identifiers
    ctrl2 <- rawData(dataset(pbmc, "ctrl"))
    ctrl2@x <- ctrl2@x + 1
    colnames(ctrl2) <- paste0(colnames(ctrl2), 2)
    pbmcNew <- optimizeNewData(pbmc, dataNew = list(ctrl2 = ctrl2),
                               useDatasets = "ctrl", nIteration = 2)
}

Description

This uses an efficient strategy for updating that takes advantage of the information in the existing factorization. It is most recommended for values of kNew smaller than current value (k, which is set when running runINMF), where it is more likely to speed up the factorization.

Usage

optimizeNewK(
  object,
  kNew,
  lambda = NULL,
  nIteration = 30,
  seed = 1,
  verbose = getOption("ligerVerbose"),
  k.new = kNew,
  max.iters = nIteration,
  rand.seed = seed,
  thresh = NULL
)

Arguments

Value

object with W slot updated with the new $W$ matrix, and the H and V slots of each ligerDataset object in the datasets slot updated with the new dataset specific $H$ and $V$ matrix, respectively.

See Also

runINMF, optimizeNewLambda, optimizeNewData

Examples

pbmc <- normalize(pbmc)
pbmc <- selectGenes(pbmc)
pbmc <- scaleNotCenter(pbmc)
# Only running a few iterations for fast examples
if (requireNamespace("RcppPlanc", quietly = TRUE)) {
    pbmc <- runINMF(pbmc, k = 20, nIteration = 2)
    pbmc <- optimizeNewK(pbmc, kNew = 25, nIteration = 2)
}

Description

Uses an efficient strategy for updating that takes advantage of the information in the existing factorization; always uses previous k. Recommended mainly when re-optimizing for higher lambda and when new lambda value is significantly different; otherwise may not return optimal results.

Usage

optimizeNewLambda(
  object,
  lambdaNew,
  nIteration = 30,
  seed = 1,
  verbose = getOption("ligerVerbose"),
  new.lambda = lambdaNew,
  max.iters = nIteration,
  rand.seed = seed,
  thresh = NULL
)

Arguments

Value

Input object with optimized factorization values updated. including the W matrix in liger object, and H and V matrices in each ligerDataset object in the datasets slot.

See Also

runINMF, optimizeNewK, optimizeNewData

Examples

pbmc <- normalize(pbmc)
pbmc <- selectGenes(pbmc)
pbmc <- scaleNotCenter(pbmc)
if (requireNamespace("RcppPlanc", quietly = TRUE)) {
    # Only running a few iterations for fast examples
    pbmc <- runINMF(pbmc, k = 20, nIteration = 2)
    pbmc <- optimizeNewLambda(pbmc, lambdaNew = 5.5, nIteration = 2)
}