R/odseq_unaligned.R

In odseq: Outlier detection in multiple sequence alignments

odseq_unaligned <- function(distance_matrix, B = 100, threshold = 0.025, type = "similarity"){
  
  n <- nrow(distance_matrix)
  
  # Compute score for each sequence
  distance_scores <- apply(distance_matrix, 1, sum)

  # Perform bootstrap of distance scores  
  
  distribution_scores <- replicate(B, {boot <- sample(distance_scores, n, replace = TRUE);
                            mean(boot)})
  
  # Return logical vector of outliers, using threshold
  
  confidence_interval <- quantile(distribution_scores, probs = c(threshold, 1 - threshold))
  
  if(type == "similarity"){
    outliers <- distance_scores < confidence_interval[1]  
  }
  else if(type == "distance"){
    outliers <- distance_scores > confidence_interval[2]
  }
  else{
    stop("Matrix type not supported. Choose similarity or distance.")
  }
  
  return(outliers)
  
}