R/decoModel.R

In BIMSBbioinfo/deconvR: Simulation and Deconvolution of Omic Profiles

#' @title A function to train user selected model to deconvolute samples
#' @param model A string indicating which model is used to deconvolute the
#' samples. Can be either "nnls" (for non-negative least squares) or "svr"
#' (support vector regression) or "qp" (quadratic programming) or "rlm" (robust
#' linear regression).
#' @param ref reference atlas provided by user in matrix form
#' @param mix bulk data provided by user in matrix form
#' @importFrom nnls nnls
#' @importFrom MASS rlm
#' @importFrom e1071 best.tune
#' @importFrom quadprog solve.QP
#' @keywords internal
#' @return the coefficient of training results
#' @noRd
decoModel <- function(model, ref, mix, colnum) {
  if (model == "nnls") { # non negative least squares
    coefficients <- (nnls::nnls(data.matrix(ref), mix))$x
  } else if (model == "svr") { # support vector regression
    model <- e1071::best.tune("svm",
      train.x = mix ~ ref, kernel = "linear",
      type = "nu-regression", scale = FALSE,
      ranges = list(nu = seq(0.25, 0.75, 0.5))
    )
    coefficients <- t(model$coefs) %*% model$SV
  } else if (model == "qp") { # quadratic programming
    Amat <- rbind(
      matrix(rep(1, colnum), nrow = 1),
      diag(colnum)
    )
    coefficients <- quadprog::solve.QP(
      Dmat = (t(ref) %*% ref), dvec = (t(ref) %*% mix),
      Amat = t(Amat), bvec = c(c(1), rep(0, colnum)),
      meq = 1
    )$solution
  } else if (model == "rlm") { # robust linear regression
    coefficients <- (MASS::rlm(ref, mix, maxit = 100))$coefficients
  }
  coefficients <- ifelse(coefficients < 0, 0, coefficients)
}