R/calculate_features.R

In ampir: Predict Antimicrobial Peptides

#' Calculate a set of numerical features from protein sequences
#'
#' This function calculates set physicochemical and compositional features from protein sequences in preparation for supervised model learning
#'
#' @note This function depends on the Peptides package
#' @references Osorio, D., Rondon-Villarreal, P. & Torres, R. Peptides: A package for data mining of antimicrobial peptides. The R Journal. 7(1), 4–14 (2015).
#'
#' @export
#'
#' @param df A dataframe which contains protein sequence names as the first column and amino acid sequence as the second column
#' @param min_len Minimum length sequence for which features can be calculated. It is an error to provide sequences with length shorter than this
#'
#' @return A dataframe containing numerical values related to the protein features of each given protein
#'
#' @examples
#'
#' my_protein_df <- read_faa(system.file("extdata/bat_protein.fasta", package = "ampir"))
#'
#' calculate_features(my_protein_df)
#' ## Output (showing the first six output columns)
#' #      seq_name     Amphiphilicity  Hydrophobicity     pI          Mw       Charge    ....
#' # [1] G1P6H5_MYOLU	   0.4145847       0.4373494     8.501312     9013.757   4.53015   ....


calculate_features <- function(df, min_len = 10) {

  df <- as.data.frame(df, stringsAsFactors = FALSE)

  short_proteins_index <- nchar(df[[2]]) < min_len
  df_cut <- df[!short_proteins_index,]
  if ( sum(short_proteins_index) > 0){
    stop("calculate_features was called on one or more sequences shorter than the specified min_len ")
  }

  seq <- df_cut[[2]]
  seq_name <- df_cut[[1]]

  Amphiphilicity    <- calc_amphiphilicity(seq)
  Hydrophobicity    <- calc_hydrophobicity(seq)
  Isoelectric_point <- calc_pI(seq)
  Mol_weight        <- calc_mw(seq)
  Net_charge        <- calc_net_charge(seq)
  Pseudo_composition<- calc_pseudo_comp(seq, lambda_min = (min_len-1))

  cbind(seq_name, Amphiphilicity, Hydrophobicity, Isoelectric_point, Mol_weight, Net_charge, Pseudo_composition, stringsAsFactors = FALSE)
}