R/groupUMI4C.R

In Pasquali-lab/UMI4Cats: UMI4Cats: Processing, analysis and visualization of UMI-4C chromatin contact data

#' Add grouping of UMI-4C samples
#' 
#' This function can be used to add specific groupings to \code{UMI4C} objects.
#' @param umi4c \linkS4class{UMI4C} object as generated by \code{\link{makeUMI4C}}.
#' @inheritParams makeUMI4C
#' @return Adds a new \code{UMI4C} object into the \code{groupsUMI4C} slot with 
#' samples grouped according to \code{grouping} variable.
#' @examples 
#' data("ex_ciita_umi4c")
#' 
#' ex_ciita_umi4c <- addGrouping(ex_ciita_umi4c, grouping="condition")
#' @export
addGrouping <- function(umi4c, 
                        grouping = "sampleID",
                        scales =  5:150,
                        normalized = TRUE,
                        sd = 2) {
  if(is.null(grouping)) stop("You need to provide a valid variable for grouping.")
  if(length(grouping)>1) stop("Use only one varible for grouping. You can latter add more groupings using the addGrouping() function.")
  else if(grouping %in% metadata(umi4c)$grouping) stop("This grouping is already included in your UMI4C object")
  
  metadata(umi4c)$grouping <- c(metadata(umi4c)$grouping, grouping)
  
  # Group matrix
  if (grouping!="sampleID") {
    umi_group <- groupSamplesUMI4C(umi4c,
                                   grouping=grouping)
  } else {
    umi_group <- umi4c
  }
  
  ## Get reference samples
  ref <- metadata(umi_group)$ref_umi4c
  
  if (is.null(ref) | !(grouping %in% names(ref))) {
    # Get sample with less UMIs if no ref present
    metadata(umi_group)$ref_umi4c <- colnames(assay(umi_group))[which(colSums(assay(umi_group)) == min(colSums(assay(umi_group))))][1]
  } else {
    # Use value from named list
    metadata(umi_group)$ref_umi4c <- ref[grouping]
  }
  
  # Get normalization matrix
  umi_group <- getNormalizationMatrix(umi_group)
  
  ## Calculate domainograms
  umi_group <- calculateDomainogram(umi_group,
                                scales = scales,
                                normalized = normalized
  )
  
  ## Calculate adaptative trend
  umi_group <- calculateAdaptativeTrend(umi_group,
                                    sd = sd,
                                    normalized = normalized
  )
  
  grouping_list <- groupsUMI4C(umi4c)
  grouping_list[[grouping]] <- umi_group
  
  groupsUMI4C(umi4c) <- grouping_list
  
  return(umi4c)
}

#' Group UMI4C samples
#' 
#' Combines UMI4C samples by adding UMIs from \code{assay(umi4c)} to represent
#' the levels in \code{grouping}.
#' @inheritParams addGrouping
#' @return A grouped \code{UMI4C} object.
groupSamplesUMI4C <- function(umi4c, grouping="condition") {
  assay <- assay(umi4c)
  
  ## Sum UMI4C from replicates
  assay_m <- reshape2::melt(assay)
  colnames(assay_m) <- c("rowname", "sampleID", "UMIs")
  assay_m <- suppressWarnings(dplyr::left_join(
    assay_m,
    data.frame(colData(umi4c))[, unique(c("sampleID", grouping)), drop = FALSE],
    by = "sampleID"
  ))
  assay_df <- assay_m %>%
    dplyr::group_by_at(c("rowname", grouping)) %>%
    dplyr::summarise(UMIs = sum(UMIs, na.rm = TRUE)) %>%
    reshape2::dcast(stats::as.formula(paste0("rowname~", grouping)), value.var = "UMIs")
  
  assay <- as.matrix(assay_df[, -which(colnames(assay_df) == "rowname")], )
  rownames(assay) <- assay_df$rowname
  colnames(assay) <- colnames(assay_df)[-1]
  
  ## Summarize colData
  colData <- data.frame(colData(umi4c)) %>%
    dplyr::group_by_at(grouping) %>%
    dplyr::summarise_all(paste0, collapse = ", ")
  
  umi4c_grouped <- UMI4C(colData = colData,
                         rowRanges = rowRanges(umi4c),
                         metadata = metadata(umi4c),
                         assays = SimpleList(umi = assay))
  
  return(umi4c_grouped)
}