Diff of /R/write_dummy_data.R [000000] .. [409433]

Switch to side-by-side view

--- a
+++ b/R/write_dummy_data.R
@@ -0,0 +1,75 @@
+#' Write random sequences to fasta file
+#' 
+#' Create random sequences from predefined vocabulary and write to fasta file.
+#' 
+#' @param file_path Output directory; can also be a file name but only possible if \code{write_to_file_path = TRUE} and 
+#' \code{num_files = 1}).
+#' @param num_files Number of files to create.
+#' @param header Fasta header name. 
+#' @param seq_length Length of one sequence. If vector longer than 1, will randomly sample from that vector. 
+#' @param num_seq Number of sequences per file.
+#' @param fasta_name_start Beginning string of file name. Output files are named fasta_name_start + _i.fasta where i is an integer index.
+#' @param write_to_file_path Whether to write output directly to \code{file_path}, i.e. file_path is not a directory.
+#' @param prob Probability of each character in the \code{vocabulary} to be sampled. If `NULL` each character has same probability.
+#' @param vocabulary Set of characters to sample sequences from. 
+#' @examples
+#' path_output <- tempfile()
+#' dir.create(path_output)
+#' create_dummy_data(file_path = path_output,
+#'                   num_files = 3,
+#'                   seq_length = 11, 
+#'                   num_seq = 5,                   
+#'                   vocabulary = c("a", "c", "g", "t"))
+#' list.files(path_output)                
+#'    
+#' @returns None. Writes data to files.     
+#' @export
+create_dummy_data <- function(file_path,
+                              num_files,
+                              header = "header", 
+                              seq_length, 
+                              num_seq,
+                              fasta_name_start = "file",
+                              write_to_file_path = FALSE,
+                              prob = NULL,
+                              vocabulary = c("a", "c", "g", "t")) {
+  
+  if (!is.null(prob)) {
+    stopifnot(length(prob) == length(vocabulary))
+    stopifnot(abs(1 - sum(prob)) < 1e-10)
+  }
+  
+  if (write_to_file_path) {
+    stopifnot(num_files == 1)
+  }
+  
+  if (!dir.exists(file_path) & !write_to_file_path) {
+    dir.create(file_path)
+  }
+  
+  for (i in 1:num_files){
+    df <- data.frame(Header = NULL, Sequence = NULL)
+    for (j in 1:num_seq) {
+      nuc_seq <- random_seq(vocabulary = vocabulary, seq_length = seq_length, prob = prob)
+      new_row <- data.frame(Header = header, Sequence = nuc_seq)
+      df <- rbind(df, new_row)
+    }
+    df$Header <- as.character(df$Header)
+    df$Sequence <- as.character(df$Sequence)
+    if (write_to_file_path) {
+      out.file <- file_path
+    } else {
+      file_name <- paste0(fasta_name_start, "_", i, ".fasta")
+      out.file <- file.path(file_path, file_name)
+    }
+    microseq::writeFasta(fdta = dplyr::as_tibble(df), 
+                         out.file =  out.file)
+  }
+}
+
+random_seq <- function(vocabulary, seq_length, prob = NULL) {
+  if (length(seq_length) > 1) {
+    seq_length <- sample(seq_length, 1)
+  }
+  sample(vocabulary, seq_length, replace = TRUE, prob = prob) %>% paste(collapse = "")
+}