R/estimate_missing_value.R

In Dahaniel/DoOR.functions: Integrating Heterogeneous Odorant Response Data into a Common Response Model: A DoOR to the Complete Olfactome

#' Estimate the missing entries in a response data
#'
#' Estimate the missing entries in a response data
#'
#' A wrapper programe for using Pearson Correlation or k-nearest neighbors to
#' estimate the missing entries in a response matrix.
#'
#' @param data a data frame or matrix, contaning the consensus response values
#' @param nodor a numeric value, specifying the number of the selected odors
#' @param method character string, specifying the method ("PC" (Pearson's
#' coefficient) and "Knn" (k nearest neighbors)) for estimation, the default is
#' "PC".
#' @author Shouwen Ma <\email{shouwen.ma@@uni-konstanz.de}>
#' @aliases DoOREst estimate_missing_value
#' @references Kim, H., Golub, G. H. & Park, H., Missing value estimation for
#' DNA microarray gene expression data: local least squares imputation., 2005,
#' Bioinformatics, 21, 187-198
#' @export
#' @keywords math
#' @examples
#' \dontrun{
#' # load data
#' library(DoOR.data)
#' data(door_response_matrix)
#' 
#' # pick an example subset
#' subset <- door_response_matrix[1:100, 11:30]
#' 
#' # estimate missing values
#' est.data <- estimate_missing_value(data = subset, nodor = 6)
#' }
estimate_missing_value <-
  function(data, nodor, method = "PC") {
    # find where is missing in the data
    inputdata <- as.matrix(data)
    outputdata <- inputdata
    whichmissing <- which(is.na(inputdata), arr.ind = TRUE)
    dim_whichmissing <- dim(whichmissing)
    precentageMissing <-
      round(dim_whichmissing[1] / length(inputdata), 4) * 100
    message(paste(
      "There are",
      dim_whichmissing[1],
      paste("(", precentageMissing, "%)", sep = ""),
      "missing values in the given data."
    ))
    
    for (i in 1:dim_whichmissing[1]) {
      receptor <- colnames(inputdata)[whichmissing[i, 2]]
      inchikey 	 <- rownames(inputdata)[whichmissing[i, 1]]
      
      if (method == "Knn") {
        tryestmatine <- try(LLSIestKnn(
          InChIKey = inchikey,
          receptor = receptor,
          response_matrix = inputdata,
          nodor = nodor
        ),
        TRUE)
      }
      
      if (method == "PC") {
        tryestmatine <- try(LLSIestPC(
          InChIKey = inchikey,
          receptor = receptor,
          response_matrix = inputdata,
          nodor = nodor
        ),
        TRUE)
      }
      
      if (inherits(tryestmatine, "try-error")) {
        print(
          paste(
            "The odorant response of",
            receptor,
            "to",
            inchikey,
            "can not be estimated."
          )
        )
      }
      else {
        outputdata[inchikey, receptor] <- tryestmatine$estimation
      }
      if (inherits(tryestmatine, "Warning messages:")) {
        stop
      }
    }
    whichmissingOutput 	<- which(is.na(outputdata), arr.ind = TRUE)
    dim_whichmissingOutput 	<- dim(whichmissingOutput)
    precentageMissingOutput <-
      round(dim_whichmissingOutput[1] / length(outputdata), 4) * 100
    N_estimate <- dim_whichmissing[1] - dim_whichmissingOutput[1]
    message(
      paste(
        "There are",
        dim_whichmissingOutput[1],
        paste("(", precentageMissingOutput, "%)", sep = ""),
        "missing values in the output data."
      )
    )
    message(paste(N_estimate, "values were estimated."))
    return(outputdata)
  }