additional arguments for MetacellsByGroups

Package: hdWGCNA

Title: hdWGCNA

Version: 0.1.1.9004

Version: 0.1.1.9005

Authors@R: c(

    person("Sam", "Morabito", , "smorabit@uci.edu", role = c("aut", "cre"),

           comment = c(ORCID = "0000-0002-7768-4856")),

# hdWGCNA 0.1.1.9005 (2022-06-17)

## Added

- None

## Changes

- Added new arguments to `MetacellsByGroups` and `ConstructMetacells` to exclude very small groups (`min_cells`), to reach a target number of metacells (`target_metacells`), and to exclude metacells with too much overlap (`max_shared`).

# hdWGCNA 0.1.1.9004 (2022-6-13)

## Added

- None

  multi_group_name = NULL,

  return_seurat = TRUE,

  wgcna_name=NULL,

  assay=NULL,

  slot = 'data',

  ...

){

  params <- GetWGCNAParams(seurat_obj, wgcna_name)

  genes_use <- GetWGCNAGenes(seurat_obj, wgcna_name)

  modules <- GetModules(seurat_obj, wgcna_name)

  if(is.null(assay)){

    assay <- params$metacell_assay

  }

  # use metacells or whole seurat object?

  if(use_metacells){

    s_obj <- seurat_obj

  }

  # check the assay:

  if(!(assay %in% names(s_obj@assays))){

    stop("Assay not found. Check names(seurat_obj@assays) or names(GetMetacellObject(seurat_obj)@assays)")

  }

  # check that group.by is in the Seurat object & in the metacell object:

  if(!(group.by %in% colnames(s_obj@meta.data))){

    m_cell_message <- ""

  # check that the group names are actually in the group.by column:

  # subset further if multiExpr:

  if(!is.null(multi.group.by)){

    seurat_meta <- seurat_meta %>% subset(get(multi.group.by) %in% multi_group_name)

#' @param return_metacell Logical to determine if we return the metacell seurat object (TRUE), or add it to the misc in the original Seurat object (FALSE). Default to FALSE.

#' @param mode determines how to make gene expression profiles for metacells from their constituent single cells. Options are "average" or "sum".

#' @param max_shared the maximum number of cells to be shared across two metacells

#' @param target_metacells the maximum target number of metacells to construct

#' @param max_iter the maximum number of iterations in the metacells bootstrapping loop

#' @param max_shared the maximum number of cells to be shared across two metacells

#' @param verbose logical indicating whether to print additional information

#' @param wgcna_name name of the WGCNA experiment

#' @keywords scRNA-seq

#' @examples

#' ConstructMetacells

ConstructMetacells <- function(

  seurat_obj, name='agg', ident.group='seurat_clusters', k=50,

  seurat_obj, name='agg', ident.group='seurat_clusters', k=25,

  reduction='umap', assay='RNA',

  cells.use = NULL, # if we don't want to use all the cells to make metacells, good for train/test split

  slot='counts',  meta=NULL, return_metacell=FALSE,

  mode = 'average', max_shared=10, verbose=FALSE, wgcna_name=NULL

  mode = 'average', max_shared=15,

  target_metacells=1000,

  max_iter=5000,

  verbose=FALSE,

  wgcna_name=NULL

){

  if(is.null(wgcna_name)){wgcna_name <- seurat_obj@misc$active_wgcna}

  get_shared <- function(other, this_choice) {

      k2 - length(union(cell_sample[other, ], this_choice))

  }

  while (length(good_choices) > 0 & it < 5000) {

  while (length(good_choices) > 0 & length(chosen) < target_metacells & it < max_iter) {

      it <- it + 1

      choice <- sample(seq_len(length(good_choices)), size = 1,

          replace = FALSE)

      this_choice <- cell_sample[nrow(cell_sample), ]

      shared <- sapply(others, get_shared, this_choice = this_choice)

      if (max(shared) < 0.9 * k) {

      # if (max(shared) < 0.9 * k) { # old way of doing it

      if(max(shared) <= max_shared){

          chosen <- new_chosen

      }

  }

  # groups of cells to combine

  mask <- sapply(seq_len(nrow(cell_sample)), function(x) seq_len(ncol(exprs_old)) %in%

      cell_sample[x, , drop = FALSE])

  mask <- mask[,which(shared_old <= max_shared)]

  cell_sample <- cell_sample[which(shared_old <= max_shared),]

  # mask <- mask[,which(shared_old <= max_shared)]

  # cell_sample <- cell_sample[which(shared_old <= max_shared),]

  mask <- Matrix::Matrix(mask)

  # average or sum expression?

  )

  # calculate stats:

  shared <- shared[shared <= max_shared]

  # shared <- shared[shared <= max_shared]

  max_shared <- max(shared)

  median_shared <- median(shared)

  mean_shared <- mean(shared)

#' @param group.by A character vector of Seurat metadata column names representing groups for which metacells will be computed.

#' @param k Number of nearest neighbors to aggregate. Default = 50

#' @param name A string appended to resulting metalcells. Default = 'agg'

#' @param reduction A dimensionality reduction stored in the Seurat object. Default = 'umap'

#' @param reduction A dimensionality reduction stored in the Seurat object. Default = 'pca'

#' @param assay Assay to extract data for aggregation. Default = 'RNA'

#' @param slot Slot to extract data for aggregation. Default = 'data'

#' @param mode determines how to make gene expression profiles for metacells from their constituent single cells. Options are "average" or "sum".

#' @param min_cells the minimum number of cells in a particular grouping to construct metacells

#' @param max_shared the maximum number of cells to be shared across two metacells

#' @param target_metacells the maximum target number of metacells to construct

#' @param max_iter the maximum number of iterations in the metacells bootstrapping loop

#' @param verbose logical indicating whether to print additional information

#' @param wgcna_name name of the WGCNA experiment

#' @keywords scRNA-seq

#' @examples

#' MetacellsByGroups

MetacellsByGroups <- function(

  seurat_obj, group.by=c('seurat_clusters'), ident.group='seurat_clusters',

  k=50, reduction='umap', assay='RNA',

  seurat_obj, group.by=c('seurat_clusters'),

  ident.group='seurat_clusters',

  k=25, reduction='pca', assay='RNA',

  cells.use = NULL, # if we don't want to use all the cells to make metacells, good for train/test split

  slot='counts', mode = 'average', max_shared=10, verbose=FALSE, wgcna_name=NULL

  slot='counts', mode = 'average', min_cells=100,

  max_shared=15,

  target_metacells=1000,

  max_iter=5000, verbose=FALSE, wgcna_name=NULL

){

  if(is.null(wgcna_name)){wgcna_name <- seurat_obj@misc$active_wgcna}

    stop('Invalid choice for mode. Mode can be either sum or average.')

  }

  # check reduction

  if(!(reduction %in% names(seurat_obj@reductions))){

    stop(paste0("Invalid reduction (", reduction, "). Reductions in Seurat object: ", paste(names(seurat_obj@reductions), collapse=', ')))

  }

  # subset seurat object by seleted cells:

  if(!is.null(cells.use)){

    seurat_full <- seurat_obj

  # remove groups that are too small:

  # TODO: add a warning to let the user know that some groups are skipped?

  groupings <- groupings[table(seurat_obj$metacell_grouping) >= 2*k]

  groupings <- groupings[table(seurat_obj$metacell_grouping) >= min_cells]

  print(groupings)

  if(length(groupings) == 0 ){

    stop("No groups met the min_cells requirement.")

  }

  # unique meta-data for each group

  meta_df <- as.data.frame(do.call(rbind, strsplit(groupings, '#')))

  colnames(meta_df) <- group.by

    seurat_obj = seurat_list,

    name = groupings,

    meta = meta_list,

    MoreArgs = list(k=k, reduction=reduction, assay=assay, slot=slot, return_metacell=TRUE, mode=mode, max_shared=max_shared, verbose=verbose, wgcna_name=wgcna_name)

    MoreArgs = list(k=k, reduction=reduction, assay=assay, slot=slot, return_metacell=TRUE, mode=mode, max_shared=max_shared, max_iter=max_iter, target_metacells=target_metacells, verbose=verbose, wgcna_name=wgcna_name)

  )

  names(metacell_list) <- groupings

      </button>

      <span class="navbar-brand">

        <a class="navbar-link" href="https://smorabit.github.io/scWGCNA/index.html">hdWGCNA</a>

        <span class="version label label-default" data-toggle="tooltip" data-placement="bottom" title="">0.1.1.9004</span>

        <span class="version label label-default" data-toggle="tooltip" data-placement="bottom" title="">0.1.1.9005</span>

      </span>

    </div>