Package 'SiNMFiD' reference manual

Title:	Supervised iNMF informed Deconvolution
Description:	A package for completing cell type deconvolution on bulk spatial transcriptomic data utilizing multiple reference scRNA-seq datasets.
Authors:	Joshua Sodicoff [aut, cre], Yichen Wang [ctb]
Maintainer:	Joshua Sodicoff <[email protected]>
License:	MIT + file LICENSE
Version:	0.0.0.9000
Built:	2024-09-12 08:32:24 UTC
Source:	https://github.com/welch-lab/SiNMFiD

Calculate relationships between cell types

Description

Calculate relationships between cell types

Usage

analyze_gene_signatures(
  filepath,
  analysis.name,
  spatial.data.name,
  rand.seed = 123,
  cell.types.use = NULL,
  return.objs = F
)
analyze_gene_signatures(
  filepath,
  analysis.name,
  spatial.data.name,
  rand.seed = 123,
  cell.types.use = NULL,
  return.objs = F
)

Arguments

`filepath`	Path to analysis directory
`analysis.name`	String identifying the analysis
`spatial.data.name`	String identifying the spatial sample
`rand.seed`	Integer random seed
`cell.types.use`	A string of cell type labels to include in the plot, by default all cell types present
`return.objs`	Logical, whether to return a list of matrices of derived data

Value

named list of cosine similarity matrix and hierarchical clustering, if return.objs = TRUE

Calculate relationships between cell type distributions

Description

Calculate relationships between cell type distributions

Usage

analyze_spatial_correlation(
  filepath,
  analysis.name,
  spatial.data.name,
  rand.seed = 123,
  mat.use = "proportions",
  cell.types.use = NULL,
  return.objs = F
)
analyze_spatial_correlation(
  filepath,
  analysis.name,
  spatial.data.name,
  rand.seed = 123,
  mat.use = "proportions",
  cell.types.use = NULL,
  return.objs = F
)

Arguments

`filepath`	Path to analysis directory
`analysis.name`	String identifying the analysis
`spatial.data.name`	String identifying the spatial sample
`rand.seed`	Integer random seed
`mat.use`	A string, either "raw" or "proportions" referring to what version of the results to summarize
`cell.types.use`	A string of cell type labels to include in the plot, by default all cell types present
`return.objs`	Logical, whether to return a list of matrices of derived data

Value

named list of pearson correlation matrix and hierarchical clustering, if return.objs = TRUE

Calculate cell sizes with all reference data

Description

Calculate cell sizes with all reference data

Usage

calculate_cell_sizes(
  data.list,
  annotations,
  filepath,
  analysis.name,
  datasets.remove = NULL,
  plot.hist = FALSE,
  chunk = 1000
)
calculate_cell_sizes(
  data.list,
  annotations,
  filepath,
  analysis.name,
  datasets.remove = NULL,
  plot.hist = FALSE,
  chunk = 1000
)

Arguments

`data.list`	Various formats are allowed, including 1. a liger object; 2. a character vector containing file names to RDS/H5 files. 3. Named list of liger object, RDS/H5 file name, matrix/dgCMatrix. List option can have element types mixed. A liger object have to be of version older than 1.99. RDS files must contain base dense matrix or dgCMatrix supported by package "Matrix". H5 files must contain dataset processed by rliger < 1.99.
`annotations`	Named factor of all cell type assignments, should be concatenated from all datasets.
`filepath`	Path to analysis directory where output sampling needs to be stored.
`analysis.name`	String identifying the analysis, used to make up a sub-folder name.
`datasets.remove`	Character vector of datasets to be excluded from sampling if `data.list` is a liger object. Named list of dataset names for exluding datasets in liger objects passed with a list `data.list`. See `sample_single_cell` examples.
`plot.hist`	Logical, if to display and save histograms of nUMIs by cell type
`chunk`	Integer chunk size for processing sparse data stored in H5. Number of cells to load into memory per iteration. Default `1000`.

Value

Nothing is returned, but the following file will be stored to local:

"<filepath>/<analysis.name>/cell_size_histogram.pdf" - A PDF file for the histogram that shows nUMI per cell distribution for each cell type
"<filepath>/<analysis.name>/cell_size.RDS" - RDS file of a named numeric vector object, total number of counts per cell type across all datasets.

Calculate the Wasserstein distance between cell-types and genes

Description

Calculate the Wasserstein distance between cell-types and genes

Usage

calculate_wasserstein(
  filepath,
  analysis.name,
  spatial.data.name,
  rand.seed = 123,
  mat.use = "proportions",
  cell.types.use = NULL,
  genes.use = NULL,
  p = 2,
  min.samples = 1,
  return.objs = F
)
calculate_wasserstein(
  filepath,
  analysis.name,
  spatial.data.name,
  rand.seed = 123,
  mat.use = "proportions",
  cell.types.use = NULL,
  genes.use = NULL,
  p = 2,
  min.samples = 1,
  return.objs = F
)

Arguments

`filepath`	Path to analysis directory
`analysis.name`	String identifying the analysis
`spatial.data.name`	String identifying the spatial sample
`rand.seed`	Integer random seed
`mat.use`	A string, either "raw" or "proportions" referring to what version of the results to summarize
`cell.types.use`	A string of cell type labels to include in the plot, by default all cell types present
`genes.use`	A string of genes to include in a plot, by default none
`p`	The `p` exponent used for the Minkowski distance
`min.samples`	Integer value, the minimum number of samples a cell type can load on and be included in the analysis
`return.objs`	Logical, whether to return a list of matrices of derived data

Value

matrix of pairwise Wasserstein distances if return.objs = TRUE

Generate histograms of loading by cell type

Description

Generate histograms of loading by cell type

Usage

cell_type_loading_histogram(
  filepath,
  analysis.name,
  spatial.data.name,
  rand.seed = 123,
  mat.use = "proportions",
  cell.types.plot = NULL,
  print.plots = TRUE,
  bin.num = 30
)
cell_type_loading_histogram(
  filepath,
  analysis.name,
  spatial.data.name,
  rand.seed = 123,
  mat.use = "proportions",
  cell.types.plot = NULL,
  print.plots = TRUE,
  bin.num = 30
)

Arguments

`filepath`	Path to analysis directory
`analysis.name`	String identifying the analysis
`spatial.data.name`	String identifying the spatial sample
`rand.seed`	Integer random seed
`mat.use`	A string, either "raw" or "proportions" referring to what version of the results to summarize
`print.plots`	Logical, whether to display results in the plots panel
`bin.num`	Integer number of bins to use in histogram
`cell.types.use`	A string of cell type labels to include in the plot, by default all cell types present

Value

nothing

Generate gifs of cell type distributions derived from deconvolution in space

Description

Generate gifs of cell type distributions derived from deconvolution in space

Usage

generate_loading_gifs(
  filepath,
  analysis.name,
  spatial.data.name,
  rand.seed = 123,
  mat.use = "proportions",
  cell.types.plot = NULL,
  filter = NULL,
  dims = c(500, 500)
)
generate_loading_gifs(
  filepath,
  analysis.name,
  spatial.data.name,
  rand.seed = 123,
  mat.use = "proportions",
  cell.types.plot = NULL,
  filter = NULL,
  dims = c(500, 500)
)

Arguments

`filepath`	Path to analysis directory
`analysis.name`	String identifying the analysis
`spatial.data.name`	String identifying the spatial sample
`rand.seed`	Integer random seed
`mat.use`	A string, either "raw" or "proportions" referring to what version of the results to summarize
`cell.types.plot`	A character vector of cell types to plot
`dims`	Integer vector of length 2 corresponding to the width and height of the RGL window

Value

nothing

Load data from one of multiple formats

Description

Load data from one of multiple formats

Usage

load_objs(objs, datasets.remove)
load_objs(objs, datasets.remove)

Arguments

objs

A named list of matrices (dgCMatrix), RDS file paths to matirces, H5 file paths to LIGER analyzed datasets.

Value

list object. List element type depends on input.

Plot results of `analyze_gene_signatures`

Description

Plot results of analyze_gene_signatures

Usage

plot_analyze_gene_signatures(
  filepath,
  analysis.name,
  spatial.data.name,
  rand.seed = 123,
  print.plots = T
)
plot_analyze_gene_signatures(
  filepath,
  analysis.name,
  spatial.data.name,
  rand.seed = 123,
  print.plots = T
)

Arguments

`filepath`	Path to analysis directory
`analysis.name`	String identifying the analysis
`spatial.data.name`	String identifying the spatial sample
`rand.seed`	Integer random seed
`print.plots`	Logical, whether to display results in the plots panel

Value

nothing

Plot results of `analyze_spatial_correlation`

Description

Plot results of analyze_spatial_correlation

Usage

plot_analyze_spatial_correlation(
  filepath,
  analysis.name,
  spatial.data.name,
  rand.seed = 123,
  print.plots = TRUE
)
plot_analyze_spatial_correlation(
  filepath,
  analysis.name,
  spatial.data.name,
  rand.seed = 123,
  print.plots = TRUE
)

Arguments

`filepath`	Path to analysis directory
`analysis.name`	String identifying the analysis
`spatial.data.name`	String identifying the spatial sample
`rand.seed`	Integer random seed
`print.plots`	Logical, whether to display results in the plots panel

Value

nothing

Plot results of `calculate_wasserstein`

Description

Plot results of calculate_wasserstein

Usage

plot_calculate_wasserstein(
  filepath,
  analysis.name,
  spatial.data.name,
  rand.seed = 123,
  print.plots = T
)
plot_calculate_wasserstein(
  filepath,
  analysis.name,
  spatial.data.name,
  rand.seed = 123,
  print.plots = T
)

Arguments

`filepath`	Path to analysis directory
`analysis.name`	String identifying the analysis
`spatial.data.name`	String identifying the spatial sample
`rand.seed`	Integer random seed
`print.plots`	Logical, whether to display results in the plots panel

Value

nothing

Plot results of `summarize_by_layer`

Description

Plot results of summarize_by_layer

Usage

plot_summarize_by_layer(
  filepath,
  analysis.name,
  spatial.data.name,
  rand.seed = 123,
  print.plots = T
)
plot_summarize_by_layer(
  filepath,
  analysis.name,
  spatial.data.name,
  rand.seed = 123,
  print.plots = T
)

Arguments

`filepath`	Path to analysis directory
`analysis.name`	String identifying the analysis
`spatial.data.name`	String identifying the spatial sample
`rand.seed`	Integer random seed
`print.plots`	Logical, whether to display results in the plots panel

Value

nothing

Generate silouhettes of the data along all three axes

Description

Generate silouhettes of the data along all three axes

Usage

reference_3d_coordinates(
  filepath,
  analysis.name,
  spatial.data.name,
  save.plots = FALSE
)
reference_3d_coordinates(
  filepath,
  analysis.name,
  spatial.data.name,
  save.plots = FALSE
)

Arguments

`filepath`	Path to analysis directory
`analysis.name`	String identifying the analysis
`spatial.data.name`	String identifying the spatial sample
`save.plots`	A logical, corresponding with if to save requested plots upon generation

Value

nothing

Transfer labels from coarse-grained sampled

Description

Transfer labels from coarse-grained sampled

Usage

register_voxel_to_label(
  filepath,
  analysis.name,
  spatial.data.name,
  labels.use,
  label.name
)
register_voxel_to_label(
  filepath,
  analysis.name,
  spatial.data.name,
  labels.use,
  label.name
)

Arguments

`filepath`	Path to analysis directory
`analysis.name`	String identifying the analysis
`spatial.data.name`	String identifying the spatial sample
`labels.use`	Named vector of labels for the prevoxelized data
`label.name`	String identifying the label set

Value

nothing

Sample from single cell reference datasets

Description

Sample from single cell reference datasets

Usage

sample_single_cell(
  data.list,
  annotations,
  filepath,
  analysis.name,
  datasets.remove = NULL,
  n.cells = 500,
  rand.seed = 123,
  chunk = 1000
)
sample_single_cell(
  data.list,
  annotations,
  filepath,
  analysis.name,
  datasets.remove = NULL,
  n.cells = 500,
  rand.seed = 123,
  chunk = 1000
)

Arguments

`data.list`	Various formats are allowed, including 1. a liger object; 2. a character vector containing file names to RDS/H5 files. 3. Named list of liger object, RDS/H5 file name, matrix/dgCMatrix. List option can have element types mixed. A liger object have to be of version older than 1.99. RDS files must contain base dense matrix or dgCMatrix supported by package "Matrix". H5 files must contain dataset processed by rliger < 1.99.
`annotations`	Named factor of cell type assignments.
`filepath`	Path to analysis directory where output sampling needs to be stored.
`analysis.name`	String identifying the analysis, used to make up a sub-folder name.
`datasets.remove`	Character vector of datasets to be excluded from sampling if `data.list` is a liger object. Named list of dataset names for exluding datasets in liger objects passed with a list `data.list`.
`n.cells`	Integer value corresponding to maximum number of samples per cell type. Default `500`.
`rand.seed`	Integer random seed for reproducible sampling.
`chunk`	Integer chunk size for processing sparse data stored in H5. Number of cells to load into memory per iteration. Default `1000`.

Value

Nothing is returned. File "norm_data.RDS" will be stored under "<filepath>/<analysis.name>/<rand.seed>/", containing a list of downsampled scaled (not centered) data matrix. File "sampled_cells.RDS" is stored at the same path, containing barcode vector of the sampled cells. File "source_annotations.RDS" is stored at "<filepath>/<analysis.name>/" which contains input annotations.

Examples

## Not run: 
# Explanation for how `datasets.remove` works with example:

names([email protected])
# above should show "data1", "data2", "data3", ...
# Then when sampling from `lig`, the first two datasets can be excluded with
sample_single_cell(data.list = lig, datasets.remove = c("data1", "data2"))

# If we got a list of liger object
sample_single_cell(data.list = list(human = lig1, mouse = lig2),
                   datasets.remove = list(human = c("data1", "data2"),
                                          mouse = c("10x1")))

## End(Not run)
## Not run: 
# Explanation for how `datasets.remove` works with example:

names(lig@raw.data)
# above should show "data1", "data2", "data3", ...
# Then when sampling from `lig`, the first two datasets can be excluded with
sample_single_cell(data.list = lig, datasets.remove = c("data1", "data2"))

# If we got a list of liger object
sample_single_cell(data.list = list(human = lig1, mouse = lig2),
                   datasets.remove = list(human = c("data1", "data2"),
                                          mouse = c("10x1")))

## End(Not run)

Add a new spatial dataset to the analysis directory

Description

Add a new spatial dataset to the analysis directory

Usage

save_spatial_data(
  filepath,
  analysis.name,
  spatial.data.file,
  coords.file,
  spatial.data.name
)
save_spatial_data(
  filepath,
  analysis.name,
  spatial.data.file,
  coords.file,
  spatial.data.name
)

Arguments

`filepath`	Path to analysis directory
`analysis.name`	String identifying the analysis
`spatial.data.file`	Path to an RDS file containing desired expression data
`coords.file`	Path to an RDS file containing desired coordinate data
`spatial.data.name`	String identifying the spatial sample

Value

nothing

select variable genes with the Kruskal-Wallis test

Description

select variable genes with the Kruskal-Wallis test

Usage

select_defining_genes(
  filepath,
  analysis.name,
  deconv.gene.num = 2000,
  gene.num.tol = 50,
  rand.seed = 123
)
select_defining_genes(
  filepath,
  analysis.name,
  deconv.gene.num = 2000,
  gene.num.tol = 50,
  rand.seed = 123
)

Arguments

`filepath`	Path to analysis directory
`analysis.name`	String identifying the analysis
`deconv.gene.num`	Integer, the number of genes to select
`gene.num.tol`	Integer, the maximum difference between the number of genes selected and `deconv.gene.num`
`rand.seed`	Integer random seed

Value

nothing

Set up new analysis directory

Description

Set up new analysis directory

Usage

start_analysis(filepath, analysis.name)
start_analysis(filepath, analysis.name)

Arguments

`filepath`	Path to analysis directory
`analysis.name`	String identifying the analysis

Value

nothing

Subset a spatial dataset by coordinates for analysis

Description

Subset a spatial dataset by coordinates for analysis

Usage

subset_spatial_data(
  filepath,
  analysis.name,
  spatial.data.name,
  subset.specs = list(c(NaN, NaN), c(NaN, NaN), c(NaN, NaN)),
  new.spatial.data.name = NULL,
  out.filepath = NULL
)
subset_spatial_data(
  filepath,
  analysis.name,
  spatial.data.name,
  subset.specs = list(c(NaN, NaN), c(NaN, NaN), c(NaN, NaN)),
  new.spatial.data.name = NULL,
  out.filepath = NULL
)

Arguments

`filepath`	Path to analysis directory
`analysis.name`	String identifying the analysis
`spatial.data.name`	String identifying the spatial sample
`subset.specs`	A list with length equal to the number of axes, with each entry a vector of length two, with the first element being the minimum value to include and the second being the maximum, or NaN to indicate a missing value
`new.spatial.data.name`	String, optional name for new analysis, otherwise the default "`spatial.data.name`subset`n.samples`" is used
`out.filepath`	Path to directory to save subset data to if not within the analysis

Value

nothing

Summarize cell-type and gene expression data by

Description

Summarize cell-type and gene expression data by

Usage

summarize_by_layer(
  filepath,
  analysis.name,
  spatial.data.name,
  rand.seed = 123,
  layer.list,
  type = "mean",
  mat.use = "proportions",
  cell.types.use = NULL,
  genes.use = NULL,
  return.objs = FALSE
)
summarize_by_layer(
  filepath,
  analysis.name,
  spatial.data.name,
  rand.seed = 123,
  layer.list,
  type = "mean",
  mat.use = "proportions",
  cell.types.use = NULL,
  genes.use = NULL,
  return.objs = FALSE
)

Arguments

`filepath`	Path to analysis directory
`analysis.name`	String identifying the analysis
`spatial.data.name`	String identifying the spatial sample
`rand.seed`	Integer random seed
`layer.list`	A named list of spatial samples by layer of interest
`type`	A string, either "mean" or "sum", how results should be combined for summary
`mat.use`	A string, either "raw", "proportions", or "assignments" referring to what version of the results to summarize
`cell.types.use`	A string of cell type labels to include in the plot, by default all cell types present
`genes.use`	A string of genes to include in a plot, by default none
`return.objs`	Logical, whether to return a list of matrices of derived data

Value

cell-type and gene expression data summarized by layer in a named list, if return.objs = TRUE

Summarize cell types present in the source annotations

Description

Summarize cell types present in the source annotations

Usage

summarize_clusters(filepath, analysis.name, return.objs = F)
summarize_clusters(filepath, analysis.name, return.objs = F)

Arguments

`filepath`	Path to analysis directory
`analysis.name`	String identifying the analysis
`return.objs`	Logical, whether to return a vector of the names of clusters

Value

A vector of unique clusters in the source annotations, if return.objs = TRUE

Summarize subregions of a vector of regions of interest

Description

Summarize subregions of a vector of regions of interest

Usage

summarize_subregions(
  regions,
  ontology.file = "Downloads/allen_structure_ontology.csv",
  return.objs = F
)
summarize_subregions(
  regions,
  ontology.file = "Downloads/allen_structure_ontology.csv",
  return.objs = F
)

Arguments

`regions`	A vector of region names
`ontology.file`	A csv describing the Allen structure ontology
`return.objs`	Logical, whether to return acronyms for all subregions found

Value

A vector of unique subregions within the provided regions, if return.objs = TRUE

Use predefined transformations to match some modalities to the Allen CCF

Description

Use predefined transformations to match some modalities to the Allen CCF

Usage

transform_coords_to_ccf(filepath, analysis.name, spatial.data.name, ish = T)
transform_coords_to_ccf(filepath, analysis.name, spatial.data.name, ish = T)

Arguments

`filepath`	Path to analysis directory
`analysis.name`	String identifying the analysis
`spatial.data.name`	String identifying the spatial sample
`ish`	Logical, if the data comes from the Allen Institute quantified ISH dataset

Value

nothing

Title

Description

Title

Usage

view_in_rgl(
  filepath,
  analysis.name,
  spatial.data.name,
  rand.seed = 123,
  cell.type,
  mat.use = "proportions",
  filter.samples = NULL,
  dims = c(500, 500)
)
view_in_rgl(
  filepath,
  analysis.name,
  spatial.data.name,
  rand.seed = 123,
  cell.type,
  mat.use = "proportions",
  filter.samples = NULL,
  dims = c(500, 500)
)

Arguments

`filepath`	Path to analysis directory
`analysis.name`	String identifying the analysis
`spatial.data.name`	String identifying the spatial sample
`rand.seed`	Integer random seed
`cell.type`	A string corresponding to one cell type found in the deconvolution results
`mat.use`	A string, either "raw" or "proportions" referring to what version of the results to summarize
`filter.samples`	Value for binarizing results, either presence above the provided threshold or absence below
`dims`	Integer vector of length 2 corresponding to the width and height of the RGL window

Value

nothing

Coarse-grain spatial data to a predetermined resolution

Description

Coarse-grain spatial data to a predetermined resolution

Usage

voxelize_single_cells(
  filepath,
  analysis.name,
  spatial.data.name,
  voxel.size,
  out.filepath = NULL,
  verbose = TRUE
)
voxelize_single_cells(
  filepath,
  analysis.name,
  spatial.data.name,
  voxel.size,
  out.filepath = NULL,
  verbose = TRUE
)

Arguments

`filepath`	Path to analysis directory
`analysis.name`	String identifying the analysis
`spatial.data.name`	String identifying the spatial sample
`voxel.size`	Integer, side length of one voxel
`out.filepath`	Path to directory to save subset data to if not within the analysis
`verbose`	Logical, if to print several lines of metadata on results

Value

nothing

`filepath`	Path to analysis directory
`analysis.name`	String identifying the analysis
`spatial.data.name`	String identifying the spatial sample
`rand.seed`	Integer random seed
`lambda`	Double, regularization parameter for which increasing penalizes dataset-specific effects
`thresh`	Double, minimum fractional change in objective function to continue iteration
`max.iters`	Integer maximum of iterations to complete before pausing
`nrep`	Number of random starts to complete
`print.obj`	Logical, if to print current value of objective
`verbose`	Logical, if to print the final objective and best random seed

`filepath`	filepath
`analysis.name`	analysis.name
`spatial.data.name`	spatial.data.name
`rand.seed`	rand.seed
`mat.use`	mat.use
`cell.types.plot`	cell.types.plot
`subregions.plot`	subregions.plot
`filter`	filter
`dims`	dims

Package 'SiNMFiD'

Help Index

Calculate relationships between cell types

Description

Usage

Arguments

Value

Calculate relationships between cell type distributions

Description

Usage

Arguments

Value

Calculate cell sizes with all reference data

Description

Usage

Arguments

Value

Calculate the Wasserstein distance between cell-types and genes

Description

Usage

Arguments

Value

Generate histograms of loading by cell type

Description

Usage

Arguments

Value

Title

Description

Usage

Arguments

Value

Title

Description

Usage

Arguments

Value

Generate gifs of cell type distributions derived from deconvolution in space

Description

Usage

Arguments

Value

Title

Description

Usage

Arguments

Value

Load data from one of multiple formats

Description

Usage

Arguments

Value

Flip axes in spatial data

Description

Usage

Arguments

Value

Title

Description

Usage

Arguments

Value

Plot results of analyze_gene_signatures

Description

Usage

Arguments

Value

Plot results of analyze_spatial_correlation

Description

Usage

Arguments

Value

Plot results of calculate_wasserstein

Description

Usage

Arguments

Value

Plot results of summarize_by_layer

Description

Usage

Plot results of `analyze_gene_signatures`

Plot results of `analyze_spatial_correlation`

Plot results of `calculate_wasserstein`

Plot results of `summarize_by_layer`