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(tfile, tempFile, method, kmerSize, gapOpen, gapExtend, match, misMatch), num(n)  {}
  15 /**************************************************************************************************/
  16 string Knn::getTaxonomy(Sequence* seq) {
  17         try {
  18                 string tax;
  19
  20                 //use database to find closest seq
  21
  22                 vector<int> closest = database->findClosestSequences(seq, num);
  23
  24                 vector<string> closestNames;
  25                 for (int i = 0; i < closest.size(); i++) {
  26                         //find that sequences taxonomy in map
  27                         it = taxonomy.find(names[closest[i]]);
  28
  29                         //is this sequence in the taxonomy file
  30                         if (it == taxonomy.end()) { //error not in file
  31                                 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();
  32                         }else{   closestNames.push_back(it->first);     }
  33                 }
  34
  35                 if (closestNames.size() == 0) {
  36                         m->mothurOut("Error: All the matches for sequence " + seq->getName() + " have been eliminated. " + seq->getName() + " will be disregarded."); m->mothurOutEndLine();
  37                         tax = "bad seq";
  38                 }else{
  39                         tax = findCommonTaxonomy(closestNames);
  40                         if (tax == "") { m->mothurOut("There are no common levels for sequence " + seq->getName() + ". " + seq->getName() + " will be disregarded."); m->mothurOutEndLine(); tax = "bad seq"; }
  41                 }
  42
  43                 simpleTax = tax;
  44                 return tax;
  45         }
  46         catch(exception& e) {
  47                 m->errorOut(e, "Knn", "getTaxonomy");
  48                 exit(1);
  49         }
  50 }
  51 /**************************************************************************************************/
  52 string Knn::findCommonTaxonomy(vector<string> closest)  {
  53         try {
  54                 vector< vector<string> > taxons;  //taxon[0] = vector of taxonomy info for closest[0].
  55                                                                                 //so if closest[0] taxonomy is Bacteria;Alphaproteobacteria;Rhizobiales;Azorhizobium_et_rel.;Methylobacterium_et_rel.;Bosea;
  56                                                                                 //taxon[0][0] = Bacteria, taxon[0][1] = Alphaproteobacteria....
  57
  58                 taxons.resize(closest.size());
  59                 int smallest = 100;
  60
  61                 for (int i = 0; i < closest.size(); i++) {
  62
  63                         string tax = taxonomy[closest[i]];  //we know its there since we checked in getTaxonomy
  64
  65                         taxons[i] = parseTax(tax);
  66
  67                         //figure out who has the shortest taxonomy info. so you can start comparing there
  68                         if (taxons[i].size() < smallest) {
  69                                 smallest = taxons[i].size();
  70                         }
  71                 }
  72
  73                 //start at the highest level all the closest seqs have
  74                 string common = "";
  75                 for (int i = (smallest-1); i >= 0; i--) {
  76
  77                         string thistax = taxons[0][i];
  78                         int num = 0;
  79                         for (int j = 1; j < taxons.size(); j++) {
  80                                 if (taxons[j][i] != thistax) { break; }
  81                                 num = j;
  82                         }
  83
  84                         if (num == (taxons.size()-1)) { //they all match at this level
  85                                 for (int k = 0; k <= i; k++) {
  86                                         common += taxons[0][k] + ';';
  87                                 }
  88                                 break;
  89                         }
  90                 }
  91
  92                 return common;
  93         }
  94         catch(exception& e) {
  95                 m->errorOut(e, "Knn", "findCommonTaxonomy");
  96                 exit(1);
  97         }
  98 }
  99 /**************************************************************************************************/
 100