knn.cpp

   1 /*
   2  *  knn.cpp
   3  *  Mothur
   4  *
   5  *  Created by westcott on 11/4/09.
   6  *  Copyright 2009 Schloss Lab. All rights reserved.
   7  *
   8  */
   9
  10 #include "knn.h"
  11
  12 /**************************************************************************************************/
  13 Knn::Knn(string tfile, string tempFile, string method, int kmerSize, float gapOpen, float gapExtend, float match, float misMatch, int n)
  14 : Classify(), num(n)  {
  15         try {
  16                 //create search database and names vector
  17                 generateDatabaseAndNames(tfile, tempFile, method, kmerSize, gapOpen, gapExtend, match, misMatch);
  18         }
  19         catch(exception& e) {
  20                 m->errorOut(e, "Knn", "Knn");
  21                 exit(1);
  22         }
  23 }
  24 /**************************************************************************************************/
  25 Knn::~Knn() {
  26         try {
  27                  delete phyloTree;
  28                  if (database != NULL) {  delete database; }
  29         }
  30         catch(exception& e) {
  31                 m->errorOut(e, "Knn", "~Knn");
  32                 exit(1);
  33         }
  34 }
  35 /**************************************************************************************************/
  36 string Knn::getTaxonomy(Sequence* seq) {
  37         try {
  38                 string tax;
  39
  40                 //use database to find closest seq
  41                 vector<int> closest = database->findClosestSequences(seq, num);
  42
  43                 if (m->control_pressed) { return tax; }
  44
  45                 vector<string> closestNames;
  46                 for (int i = 0; i < closest.size(); i++) {
  47                         //find that sequences taxonomy in map
  48                         it = taxonomy.find(names[closest[i]]);
  49
  50                         //is this sequence in the taxonomy file
  51                         if (it == taxonomy.end()) { //error not in file
  52                                 m->mothurOut("Error: sequence " + names[closest[i]] + " is not in the taxonomy file.  It will be eliminated as a match to sequence " + seq->getName() + "."); m->mothurOutEndLine();
  53                         }else{   closestNames.push_back(it->first);     }
  54                 }
  55
  56                 if (closestNames.size() == 0) {
  57                         m->mothurOut("Error: All the matches for sequence " + seq->getName() + " have been eliminated. " + seq->getName() + " will be disregarded."); m->mothurOutEndLine();
  58                         tax = "bad seq";
  59                 }else{
  60                         tax = findCommonTaxonomy(closestNames);
  61                         if (tax == "") { m->mothurOut("There are no common levels for sequence " + seq->getName() + ". " + seq->getName() + " will be disregarded."); m->mothurOutEndLine(); tax = "bad seq"; }
  62                 }
  63
  64                 simpleTax = tax;
  65                 return tax;
  66         }
  67         catch(exception& e) {
  68                 m->errorOut(e, "Knn", "getTaxonomy");
  69                 exit(1);
  70         }
  71 }
  72 /**************************************************************************************************/
  73 string Knn::findCommonTaxonomy(vector<string> closest)  {
  74         try {
  75                 vector< vector<string> > taxons;  //taxon[0] = vector of taxonomy info for closest[0].
  76                                                                                 //so if closest[0] taxonomy is Bacteria;Alphaproteobacteria;Rhizobiales;Azorhizobium_et_rel.;Methylobacterium_et_rel.;Bosea;
  77                                                                                 //taxon[0][0] = Bacteria, taxon[0][1] = Alphaproteobacteria....
  78
  79                 taxons.resize(closest.size());
  80                 int smallest = 100;
  81
  82                 for (int i = 0; i < closest.size(); i++) {
  83                         if (m->control_pressed) { return "control"; }
  84
  85                         string tax = taxonomy[closest[i]];  //we know its there since we checked in getTaxonomy
  86
  87                         taxons[i] = parseTax(tax);
  88
  89                         //figure out who has the shortest taxonomy info. so you can start comparing there
  90                         if (taxons[i].size() < smallest) {
  91                                 smallest = taxons[i].size();
  92                         }
  93                 }
  94
  95                 //start at the highest level all the closest seqs have
  96                 string common = "";
  97                 for (int i = (smallest-1); i >= 0; i--) {
  98                         if (m->control_pressed) { return "control"; }
  99
 100                         string thistax = taxons[0][i];
 101                         int num = 0;
 102                         for (int j = 1; j < taxons.size(); j++) {
 103                                 if (taxons[j][i] != thistax) { break; }
 104                                 num = j;
 105                         }
 106
 107                         if (num == (taxons.size()-1)) { //they all match at this level
 108                                 for (int k = 0; k <= i; k++) {
 109                                         common += taxons[0][k] + ';';
 110                                 }
 111                                 break;
 112                         }
 113                 }
 114
 115                 return common;
 116         }
 117         catch(exception& e) {
 118                 m->errorOut(e, "Knn", "findCommonTaxonomy");
 119                 exit(1);
 120         }
 121 }
 122 /**************************************************************************************************/
 123