R/get_normalized.R

In YosefLab/scone: Single Cell Overview of Normalized Expression data

#' @describeIn get_normalized 
#'   If 
#'   \code{method} is a character, it will return the normalized
#'   matrix corresponding to the normalization scheme specified
#'   by the character string.The string must be one of the 
#'   \code{row.names} of the slot \code{scone_params}.
#'   
#' @importFrom rhdf5 h5ls h5read
#' @export
#' 
#' @param log logical. Should the data be returned in log-scale
#'   
setMethod(
  f = "get_normalized",
  signature = signature(x = "SconeExperiment", method = "character"),
  definition =  function(x, method, log=FALSE) {

    if(!(method %in% rownames(x@scone_params))) {
      stop("`method` must be one of the row.names of the slot scone_params.")
    }

    if(x@scone_run == "in_memory") {
      retval <- assay(x, method)
    }

    if(x@scone_run == "hdf5") {
      retval <- h5read(x@hdf5_pointer, method)
      rownames(retval) <- h5read(x@hdf5_pointer, "genes")
      colnames(retval) <- h5read(x@hdf5_pointer, "samples")
    }

    if(x@scone_run == "no") {
      params <- unlist(x@scone_params[method,])

      if(length(x@impute_args) > 0) {
        impute_args <- x@impute_args
      } else {
        impute_args <- NULL
      }

      imputed <- x@imputation_fn[[params[1]]](assay(x), impute_args)
      scaled <- x@scaling_fn[[params[2]]](imputed)

      if(x@rezero) {
        toz = assay(x) <= 0
        scaled[toz] = 0
      }

      ruv_factors <- qc_factors <- NULL
      if(params[3] != "no_uv") {

        k <- as.numeric(strsplit(params[3], "=")[[1]][2])

        if(grepl("ruv", params[3])) {
          r <- RUVg(log1p(scaled), get_negconruv(x), k, isLog=TRUE)$W
          ruv_factors <- list(r)
          names(ruv_factors) <- paste(params[1:2], collapse="_")
        }

        if(grepl("qc", params[3])) {
          qc_factors <- prcomp(get_qc(x), center=TRUE, scale=TRUE)$x
        }

      }

      parsed <- .parse_row(params, get_bio(x), 
                           get_batch(x), ruv_factors, qc_factors)
      design_mat <- make_design(parsed$bio, parsed$batch, parsed$W,
                                nested=(x@nested &
                                          !is.null(parsed$bio) &
                                          !is.null(parsed$batch)))
      adjusted <- lm_adjust(log1p(scaled), design_mat, get_batch(x))
      
      if(x@fixzero) {
        toz = assay(x) <= 0
        adjusted[toz] = 0
        adjusted[adjusted <= 0] = 0
      }
      
      retval <- expm1(adjusted)
    }

    if(log) {
      retval <- log1p(retval)
    }

    return(retval)
  }
)


#' @describeIn get_normalized 
#'   If 
#'   \code{method} is a numeric, it will return the normalized
#'   matrix according to the scone ranking.
#'   
#' @details The numeric method will always return the normalization 
#'   corresponding to row \code{method} of the \code{scone_params} slot. This
#'   means that if \code{\link{scone}} was run with \code{eval=TRUE}, 
#'   \code{get_normalized(x, 1)} will return the top ranked method. If 
#'   \code{\link{scone}} was run with \code{eval=FALSE}, 
#'   \code{get_normalized(x,1)} will return the first normalization
#'   in the order saved by scone.
#'   
#' @export
#' 
setMethod(
  f = "get_normalized",
  signature = signature(x = "SconeExperiment", method = "numeric"),
  definition =  function(x, method, log=FALSE) {
    norm_method <- rownames(x@scone_params)[method]

    return(get_normalized(x, norm_method, log))
  }
)