R/desc-15-PSSMFeature.R

In nanxstats/protr: Generating Various Numerical Representation Schemes for Protein Sequences

#' Profile-based protein representation derived by PSSM
#' (Position-Specific Scoring Matrix)
#'
#' This function calculates the profile-based protein representation
#' derived by PSSM. The feature vector is based on the PSSM computed by
#' \code{\link{extractPSSM}}.
#'
#' For a given sequence, the PSSM feature represents the log-likelihood
#' of the substitution of the 20 types of amino acids at that position
#' in the sequence.
#'
#' Each PSSM feature value in the vector represents the degree of
#' conservation of a given amino acid type. The value is normalized
#' to interval (0, 1) by the transformation 1/(1+e^(-x)).
#'
#' @param pssmmat The PSSM computed by \code{\link{extractPSSM}}.
#'
#' @return A numeric vector which has \code{20 x N} named elements,
#' where \code{N} is the size of the window (number of rows of the PSSM).
#'
#' @seealso \link{extractPSSM} \link{extractPSSMAcc}
#'
#' @author Nan Xiao <\url{https://nanx.me}>
#'
#' @export extractPSSMFeature
#'
#' @references
#' Ye, Xugang, Guoli Wang, and Stephen F. Altschul.
#' "An assessment of substitution scores for protein profile-profile comparison."
#' \emph{Bioinformatics} 27.24 (2011): 3356--3363.
#'
#' Rangwala, Huzefa, and George Karypis.
#' "Profile-based direct kernels for remote homology detection and fold recognition."
#' \emph{Bioinformatics} 21.23 (2005): 4239--4247.
#'
#' @examples
#' if (Sys.which("makeblastdb") == "" | Sys.which("psiblast") == "") {
#'   cat("Cannot find makeblastdb or psiblast. Please install NCBI Blast+")
#' } else {
#'   x <- readFASTA(system.file(
#'     "protseq/P00750.fasta",
#'     package = "protr"
#'   ))[[1]]
#'   dbpath <- tempfile("tempdb", fileext = ".fasta")
#'   invisible(file.copy(from = system.file(
#'     "protseq/Plasminogen.fasta",
#'     package = "protr"
#'   ), to = dbpath))
#'
#'   pssmmat <- extractPSSM(seq = x, database.path = dbpath)
#'   pssmfeature <- extractPSSMFeature(pssmmat)
#'   head(pssmfeature)
#' }
extractPSSMFeature <- function(pssmmat) {

  # Normalize PSSM scores to (0, 1)
  res <- as.vector(1 / (1 + exp(pssmmat)))

  AADict <- c(
    "A", "R", "N", "D", "C", "Q", "E", "G", "H", "I",
    "L", "K", "M", "F", "P", "S", "T", "W", "Y", "V"
  )
  names(res) <- paste0(
    "PSSM_", paste0(rep(1L:ncol(pssmmat), each = 20L), "_", AADict)
  )

  res
}