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 string Knn::getTaxonomy(Sequence* seq) {
  26         try {
  27                 string tax;
  28
  29                 //use database to find closest seq
  30                 vector<int> closest = database->findClosestSequences(seq, num);
  31
  32                 if (m->control_pressed) { return tax; }
  33
  34                 vector<string> closestNames;
  35                 for (int i = 0; i < closest.size(); i++) {
  36                         //find that sequences taxonomy in map
  37                         it = taxonomy.find(names[closest[i]]);
  38
  39                         //is this sequence in the taxonomy file
  40                         if (it == taxonomy.end()) { //error not in file
  41                                 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();
  42                         }else{   closestNames.push_back(it->first);     }
  43                 }
  44
  45                 if (closestNames.size() == 0) {
  46                         m->mothurOut("Error: All the matches for sequence " + seq->getName() + " have been eliminated. " + seq->getName() + " will be disregarded."); m->mothurOutEndLine();
  47                         tax = "bad seq";
  48                 }else{
  49                         tax = findCommonTaxonomy(closestNames);
  50                         if (tax == "") { m->mothurOut("There are no common levels for sequence " + seq->getName() + ". " + seq->getName() + " will be disregarded."); m->mothurOutEndLine(); tax = "bad seq"; }
  51                 }
  52
  53                 simpleTax = tax;
  54                 return tax;
  55         }
  56         catch(exception& e) {
  57                 m->errorOut(e, "Knn", "getTaxonomy");
  58                 exit(1);
  59         }
  60 }
  61 /**************************************************************************************************/
  62 string Knn::findCommonTaxonomy(vector<string> closest)  {
  63         try {
  64                 vector< vector<string> > taxons;  //taxon[0] = vector of taxonomy info for closest[0].
  65                                                                                 //so if closest[0] taxonomy is Bacteria;Alphaproteobacteria;Rhizobiales;Azorhizobium_et_rel.;Methylobacterium_et_rel.;Bosea;
  66                                                                                 //taxon[0][0] = Bacteria, taxon[0][1] = Alphaproteobacteria....
  67
  68                 taxons.resize(closest.size());
  69                 int smallest = 100;
  70
  71                 for (int i = 0; i < closest.size(); i++) {
  72                         if (m->control_pressed) { return "control"; }
  73
  74                         string tax = taxonomy[closest[i]];  //we know its there since we checked in getTaxonomy
  75
  76                         taxons[i] = parseTax(tax);
  77
  78                         //figure out who has the shortest taxonomy info. so you can start comparing there
  79                         if (taxons[i].size() < smallest) {
  80                                 smallest = taxons[i].size();
  81                         }
  82                 }
  83
  84                 //start at the highest level all the closest seqs have
  85                 string common = "";
  86                 for (int i = (smallest-1); i >= 0; i--) {
  87                         if (m->control_pressed) { return "control"; }
  88
  89                         string thistax = taxons[0][i];
  90                         int num = 0;
  91                         for (int j = 1; j < taxons.size(); j++) {
  92                                 if (taxons[j][i] != thistax) { break; }
  93                                 num = j;
  94                         }
  95
  96                         if (num == (taxons.size()-1)) { //they all match at this level
  97                                 for (int k = 0; k <= i; k++) {
  98                                         common += taxons[0][k] + ';';
  99                                 }
 100                                 break;
 101                         }
 102                 }
 103
 104                 return common;
 105         }
 106         catch(exception& e) {
 107                 m->errorOut(e, "Knn", "findCommonTaxonomy");
 108                 exit(1);
 109         }
 110 }
 111 /**************************************************************************************************/
 112