R/ngrams.R

In biogram: N-Gram Analysis of Biological Sequences

#' Get all possible n-Grams
#'
#' Creates the vector of all possible n_grams (for given \code{n}).
#'
#' @inheritParams count_ngrams
#' @param possible_grams number of possible n-grams. If not \code{NULL} n-grams do not
#' contain information about position
#' @return a character vector. Elements of n-gram are separated by dot.
#' @note Input data must be a matrix or data frame of numeric elements.
#' @details See Details section of \code{\link{count_ngrams}} for more 
#' information about n-grams naming convention. The possible information about distance 
#' must be added by hand (see examples).
#' @export
#' @examples 
#' # bigrams for standard aminoacids
#' create_ngrams(2, 1L:20)
#' # bigrams for standard aminoacids with positions, 10 amino acid long sequence, so 
#' # only 9 bigrams can be located in sequence
#' create_ngrams(2, 1L:20, 9)
#' # bigrams for DNA with positions, 10 nucleotide long sequence, distance 1, so only 
#' # 8 bigrams in sequence
#' # paste0 adds information about distance at the end of n-gram
#' paste0(create_ngrams(2, 1L:4, 8), "_0")


create_ngrams <- function(n, u, possible_grams = NULL) {
  grid_list <- lapply(1L:n, function(i) u)
  res <- apply(expand.grid(grid_list), 1, function(x)
    paste(x, collapse = "."))
  if (!is.null(possible_grams))
    res <- as.vector(sapply(res, function(i) paste(1L:possible_grams, i, sep = "_")))
  res
}


#' Extract n-grams from sequence
#'
#' Extracts vector of n-grams present in sequence(s).
#'
#' @inheritParams count_ngrams
#' @details A format of \code{d} vector is discussed in Details of 
#' \code{\link{count_ngrams}}.
#' @return A \code{character} matrix of n-grams, where every row corresponds to a
#' different sequence.
#' @export
#' @examples 
#' # trigrams from multiple sequences
#' seqs <- matrix(sample(1L:4, 600, replace = TRUE), ncol = 50)
#' seq2ngrams(seqs, 3, 1L:4)

seq2ngrams <- function(seq, n, u, d = 0, pos = FALSE) {
  if (!(is.matrix(seq) || is.vector(seq)))
    stop("'seq' must be vector or matrix.")
  
  # if sequence is not a matrix (single sequence), convert it to matrix with 1 row
  if (!(is.matrix(seq)))
    seq <- matrix(seq, nrow = 1)
  
  # length of sequence
  len_seq <- ncol(seq)
  # number of sequences
  n_seqs <- nrow(seq)
  
  # look for n-gram indices for d
  ngram_ind <- get_ngrams_ind(len_seq, n, d)
  
  max_grams <- calc_max_grams(len_seq, n, ngram_ind)
  
  # extract n-grams from sequene
  res <- t(vapply(1L:n_seqs, function(i) {
    grams <- seq2ngrams_helper(seq[i, ], ind = ngram_ind, max_grams)
    paste(grams, paste0(attr(ngram_ind, "d"), collapse = "."), 
          sep = "_")
  }, rep("a", max_grams)))
  if (max_grams == 1)
    res <- t(res)
  
  # add position information if requested
  if(pos)
    res <- do.call(cbind, lapply(1L:ncol(res), function(pos_id)
      paste0(pos_id, "_", res[, pos_id])))
  
  res
}