R/blast_best_hit.R

In drostlab/metablastr: Perform Massive Local BLAST Searches

#' @title Retrieve only the best BLAST hit for each query
#' @description This function performs a BLAST search between query and subject sequences and
#' returns only the best hit based on the following criteria.
#' 
#' A best blast hit is defined as:
#' \itemize{
#' \item the hit with the smallest e-value
#' \item if e-values are identical then the hit with the longest alignment length is chosen
#' }
#' @param query path to input file in fasta format.
#' @param subject path to subject file in fasta format or blast-able database.
#' @param search_type type of query and subject sequences that will be compared via BLAST search.
#' Options are:
#' \itemize{
#' \item \code{search_type = "nucleotide_to_nucleotide"}
#' \item \code{search_type = "nucleotide_to_protein"}
#' \item \code{search_type = "protein_to_nucleotide"}
#' \item \code{search_type = "protein_to_protein"}
#' }
#' @param strand Query DNA strand(s) to search against database/subject.
#' Options are:
#' \itemize{
#' \item \code{strand = "both"} (Default): query against both DNA strands.
#' \item \code{strand = "minus"} : query against minus DNA strand.
#' \item \code{strand = "plus"} : query against plus DNA strand.
#' }
#' @param output.path path to folder at which BLAST output table shall be stored. 
#' Default is \code{output.path = NULL} (hence \code{getwd()} is used).
#' @param is.subject.db logical specifying whether or not the \code{subject} file is a file in fasta format (\code{is.subject.db = FALSE}; default)
#' or a blast-able database that was formatted with \code{makeblastdb} (\code{is.subject.db = TRUE}).
#' @param task BLAST search task option (depending on the selected \code{search_type}). Options are:
#' \itemize{
#' \item \code{search_type = "nucleotide_to_nucleotide"}
#' \itemize{
#' \item \code{task = "blastn"} : Standard nucleotide-nucleotide comparisons (default) - Traditional BLASTN requiring an exact match of 11.
#' \item \code{task = "blastn-short"} : Optimized nucleotide-nucleotide comparisons for query sequences shorter than 50 nucleotides.
#' \item \code{task = "dc-megablast"} : Discontiguous megablast used to find somewhat distant sequences.
#' \item \code{task = "megablast"} : Traditional megablast used to find very similar (e.g., intraspecies or closely related species) sequences.
#' \item \code{task = "rmblastn"}
#' }
#' \item \code{search_type = "nucleotide_to_protein"}
#' \itemize{
#' \item \code{task = "blastx"} : Standard nucleotide-protein comparisons (default).
#' \item \code{task = "blastx-fast"} : Optimized nucleotide-protein comparisons.
#' }
#' \item \code{search_type = "protein_to_nucleotide"}
#' \itemize{
#' \item \code{task = "tblastn"} : Standard protein-nucleotide comparisons (default).
#' \item \code{task = "tblastn-fast"} : Optimized protein-nucleotide comparisons.
#' }
#' \item \code{search_type = "protein_to_protein"}
#' \itemize{
#' \item \code{task = "blastp"} : Standard protein-protein comparisons (default).
#' \item \code{task = "blast-fast"} : Improved BLAST searches using longer words for protein seeding.
#' \item \code{task = "blastp-short"} : Optimized protein-protein comparisons for query sequences shorter than 30 residues.
#' }
#' }
#' @param db.import shall the BLAST output be stored in a PostgresSQL database and shall a connection be established to this database? Default is \code{db.import = FALSE}.
#' In case users wish to to only generate a BLAST output file without importing it to the current R session they can specify \code{db.import = NULL}.
#' @param postgres.user when \code{db.import = TRUE} and \code{out.format = "postgres"} is selected, the BLAST output is imported and stored in a 
#' PostgresSQL database. In that case, users need to have PostgresSQL installed and initialized on their system. 
#' Please consult the Installation Vignette for details. 
#' @param evalue Expectation value (E) threshold for saving hits (default: \code{evalue = 0.001}).
#' @param out.format a character string specifying the format of the file in which the BLAST results shall be stored.
#' Available options are:
#'  \itemize{
#'  \item \code{out.format = "pair"} : Pairwise
#'  \item \code{out.format = "qa.ident"} : Query-anchored showing identities
#'  \item \code{out.format = "qa.nonident"} : Query-anchored no identities
#'  \item \code{out.format = "fq.ident"} : Flat query-anchored showing identities
#'  \item \code{out.format = "fq.nonident"} : Flat query-anchored no identities
#'  \item \code{out.format = "xml"} : XML
#'  \item \code{out.format = "tab"} : Tabular separated file
#'  \item \code{out.format = "tab.comment"} : Tabular separated file with comment lines
#'  \item \code{out.format = "ASN.1.text"} : Seqalign (Text ASN.1)
#'  \item \code{out.format = "ASN.1.binary"} : Seqalign (Binary ASN.1)
#'  \item \code{out.format = "csv"} : Comma-separated values
#'  \item \code{out.format = "ASN.1"} : BLAST archive (ASN.1)
#'  \item \code{out.format = "json.seq.aln"} : Seqalign (JSON)
#'  \item \code{out.format = "json.blast.multi"} : Multiple-file BLAST JSON
#'  \item \code{out.format = "xml2.blast.multi"} : Multiple-file BLAST XML2
#'  \item \code{out.format = "json.blast.single"} : Single-file BLAST JSON
#'  \item \code{out.format = "xml2.blast.single"} : Single-file BLAST XML2
#'  \item \code{out.format = "SAM"} : Sequence Alignment/Map (SAM)
#'  \item \code{out.format = "report"} : Organism Report
#'  }
#' @param cores number of cores for parallel BLAST searches.
#' @param max.target.seqs maximum number of aligned sequences that shall be retained. Please be aware that \code{max.target.seqs} selects best hits based on the database entry and not by the best e-value. See details here: https://academic.oup.com/bioinformatics/advance-article/doi/10.1093/bioinformatics/bty833/5106166 .
#' @param db.soft.mask shall low complexity regions be soft masked? Default is \code{db.soft.mask = FALSE}.
#' @param db.hard.mask shall low complexity regions be hard masked? Default is \code{db.hard.mask = FALSE}.
#' @param blast.path path to BLAST executables.
#' @author Hajk-Georg Drost
#' @examples 
#' \dontrun{
#' blast_best_test <- blast_best_hit(
#'                  query   = system.file('seqs/qry_nn.fa', package = 'metablastr'),
#'                  subject = system.file('seqs/sbj_nn_best_hit.fa', package = 'metablastr'),
#'                  search_type = "nucleotide_to_nucleotide",
#'                  output.path = tempdir(),
#'                  db.import  = FALSE)
#'                  
#'  # look at results
#'  blast_best_test
#' }
#' 
#' @seealso \code{\link{blast_nucleotide_to_nucleotide}}, \code{\link{blast_protein_to_protein}},
#' \code{\link{blast_nucleotide_to_protein}}, \code{\link{blast_protein_to_nucleotide}},
#' \code{\link{blast_best_reciprocal_hit}}
#' @export

blast_best_hit <-
  function(query,
           subject,
           search_type = "nucleotide_to_nucleotide",
           strand = "both",
           output.path = NULL,
           is.subject.db = FALSE,
           task = "blastn",
           db.import = FALSE,
           postgres.user = NULL,
           evalue = 0.001,
           out.format = "csv",
           cores = 1,
           max.target.seqs = 10000,
           db.soft.mask = FALSE,
           db.hard.mask = FALSE,
           blast.path = NULL) {
    
  
  if (!is.element(search_type, c("nucleotide_to_nucleotide", "nucleotide_to_protein", "protein_to_nucleotide", "protein_to_protein")))
    stop("Please provide a search_type that is supported by this function. Options are search_type = 'nucleotide_to_nucleotide'; search_type = 'nucleotide_to_protein'; ",
         "search_type = 'protein_to_nucleotide'; search_type = 'protein_to_nucleotide'", call. = FALSE)
  
  '.' <- NULL
  get_best_hit <- function(x) {
    min_val <- min(x$evalue)
    evalue <- alig_length <- NULL
    res <- dplyr::filter(x, evalue == min_val)
    if (nrow(res) > 1) {
      max_len <- max(res$alig_length)
      res <- dplyr::filter(res, alig_length == max_len)
    }
    
    if (nrow(res) > 1) 
      res <- dplyr::slice(res, 1)
    
    return(res)
  }
  
  if (search_type == "nucleotide_to_nucleotide") {
    
    blast_res <- blast_nucleotide_to_nucleotide(
      query = query,
      subject = subject,
      strand = strand,
      output.path = output.path,
      is.subject.db = is.subject.db,
      task = task,
      db.import = db.import,
      postgres.user = postgres.user,
      evalue = evalue,
      out.format = out.format,
      cores = cores,
      max.target.seqs = max.target.seqs,
      db.soft.mask = db.soft.mask,
      db.hard.mask = db.hard.mask,
      blast.path = blast.path
    )
    
    query_id <- NULL
    hit_tbl <- dplyr::do(dplyr::group_by(blast_res, query_id), get_best_hit(.))
  }
  
  if (search_type == "nucleotide_to_protein") {
    
    blast_res <- blast_nucleotide_to_protein(
      query = query,
      subject = subject,
      strand = strand,
      output.path = output.path,
      is.subject.db = is.subject.db,
      task = task,
      db.import = db.import,
      postgres.user = postgres.user,
      evalue = evalue,
      out.format = out.format,
      cores = cores,
      max.target.seqs = max.target.seqs,
      db.soft.mask = db.soft.mask,
      db.hard.mask = db.hard.mask,
      blast.path = blast.path
    )
    
    query_id <- NULL
    hit_tbl <- dplyr::do(dplyr::group_by(blast_res, query_id), get_best_hit(.))
  }
  
  if (search_type == "protein_to_nucleotide") {
    
    blast_res <- blast_protein_to_nucleotide(
      query = query,
      subject = subject,
      output.path = output.path,
      is.subject.db = is.subject.db,
      task = task,
      db.import = db.import,
      postgres.user = postgres.user,
      evalue = evalue,
      out.format = out.format,
      cores = cores,
      max.target.seqs = max.target.seqs,
      db.soft.mask = db.soft.mask,
      db.hard.mask = db.hard.mask,
      blast.path = blast.path
    )
    
    query_id <- NULL
    hit_tbl <- dplyr::do(dplyr::group_by(blast_res, query_id), get_best_hit(.))
  }
  
  if (search_type == "protein_to_protein") {
    
    blast_res <- blast_protein_to_protein(
      query = query,
      subject = subject,
      output.path = output.path,
      is.subject.db = is.subject.db,
      task = task,
      db.import = db.import,
      postgres.user = postgres.user,
      evalue = evalue,
      out.format = out.format,
      cores = cores,
      max.target.seqs = max.target.seqs,
      db.soft.mask = db.soft.mask,
      db.hard.mask = db.hard.mask,
      blast.path = blast.path
    )
    
    query_id <- NULL
    hit_tbl <- dplyr::do(dplyr::group_by(blast_res, query_id), get_best_hit(.))
  }
  
  return(hit_tbl)
}