5 * Created by Pat Schloss on 12/16/08.
6 * Copyright 2008 Patrick D. Schloss. All rights reserved.
8 * This class is a child class of the Database class, which stores the template sequences as a kmer table and provides
9 * a method of searching the kmer table for the sequence with the most kmers in common with a query sequence.
10 * kmerLocations is the primary storage variable that is a two-dimensional vector where each row represents the
11 * different number of kmers and each column contains the index to sequences that use that kmer.
13 * Construction of an object of this type will first look for an appropriately named database file and if it is found
14 * then will read in the database file (readKmerDB), otherwise it will generate one and store the data in memory
17 * The search method used here is roughly the same as that used in the SimRank program that is found at the
18 * greengenes website. The default kmer size is 7. The speed complexity is between O(L) and O(LN). When I use 7mers
19 * on average a kmer is found in ~100 other sequences with a database of ~5000 sequences. If this is the case then the
20 * time would be on the order of O(0.1LN) -> fast.
25 #include "sequence.hpp"
27 #include "database.hpp"
30 /**************************************************************************************************/
32 KmerDB::KmerDB(string fastaFileName, int kSize) : Database(fastaFileName), kmerSize(kSize) {
34 string kmerDBName = fastaFileName.substr(0,fastaFileName.find_last_of(".")+1) + char('0'+ kmerSize) + "mer";
35 ifstream kmerFileTest(kmerDBName.c_str());
37 int power4s[14] = { 1, 4, 16, 64, 256, 1024, 4096, 16384, 65536, 262144, 1048576, 4194304, 16777216, 67108864 };
39 maxKmer = power4s[kmerSize];
40 kmerLocations.resize(maxKmer+1);
42 if(!kmerFileTest){ // if we can open the kmer db file, then read it in...
43 mothurOut("Generating the " + kmerDBName + " database...\t"); cout.flush();
44 generateKmerDB(kmerDBName);
46 else{ // ...otherwise generate it.
47 mothurOut("Reading in the " + kmerDBName + " database...\t"); cout.flush();
48 readKmerDB(kmerDBName, kmerFileTest);
50 mothurOut("DONE."); mothurOutEndLine(); mothurOutEndLine(); cout.flush();
53 /**************************************************************************************************/
57 //for (int i = 0; i < templateSequences.size(); i++) { delete templateSequences[i]; }
58 // templateSequences.clear();
61 /**************************************************************************************************/
63 Sequence KmerDB::findClosestSequence(Sequence* candidateSeq){
70 vector<int> matches(numSeqs, 0); // a record of the sequences with shared kmers
71 vector<int> timesKmerFound(kmerLocations.size()+1, 0); // a record of the kmers that we have already found
73 int numKmers = candidateSeq->getNumBases() - kmerSize + 1;
75 for(int i=0;i<numKmers;i++){
76 int kmerNumber = kmer.getKmerNumber(candidateSeq->getUnaligned(), i); // go through the query sequence and get a kmer number
77 if(timesKmerFound[kmerNumber] == 0){ // if we haven't seen it before...
78 for(int j=0;j<kmerLocations[kmerNumber].size();j++){//increase the count for each sequence that also has
79 matches[kmerLocations[kmerNumber][j]]++; // that kmer
82 timesKmerFound[kmerNumber] = 1; // ok, we've seen the kmer now
85 for(int i=0;i<numSeqs;i++){ // find the db sequence that shares the most kmers with
86 if(matches[i] > searchScore){ // the query sequence
87 searchScore = matches[i];
92 searchScore = 100 * searchScore / (float) numKmers; // return the Sequence object corresponding to the db
93 return templateSequences[maxSequence]; // sequence with the most shared kmers.
96 /**************************************************************************************************/
98 void KmerDB::generateKmerDB(string kmerDBName){
102 for(int i=0;i<numSeqs;i++){ // for all of the template sequences...
104 string seq = templateSequences[i].getUnaligned(); // ...take the unaligned sequence...
105 int numKmers = seq.length() - kmerSize + 1;
107 vector<int> seenBefore(maxKmer+1,0);
108 for(int j=0;j<numKmers;j++){ // ...step though the sequence and get each kmer...
109 int kmerNumber = kmer.getKmerNumber(seq, j);
110 if(seenBefore[kmerNumber] == 0){
111 kmerLocations[kmerNumber].push_back(i); // ...insert the sequence index into kmerLocations for
112 } // the appropriate kmer number
113 seenBefore[kmerNumber] = 1;
117 ofstream kmerFile; // once we have the kmerLocations folder print it out
118 openOutputFile(kmerDBName, kmerFile); // to a file
120 for(int i=0;i<maxKmer;i++){ // step through all of the possible kmer numbers
121 kmerFile << i << ' ' << kmerLocations[i].size(); // print the kmer number and the number of sequences with
122 for(int j=0;j<kmerLocations[i].size();j++){ // that kmer. then print out the indices of the sequences
123 kmerFile << ' ' << kmerLocations[i][j]; // with that kmer.
131 /**************************************************************************************************/
133 void KmerDB::readKmerDB(string kmerDBName, ifstream& kmerDBFile){
135 kmerDBFile.seekg(0); // start at the beginning of the file
140 for(int i=0;i<maxKmer;i++){
142 kmerDBFile >> seqName >> numValues;
144 for(int j=0;j<numValues;j++){ // for each kmer number get the...
145 kmerDBFile >> seqNumber; // 1. number of sequences with the kmer number
146 kmerLocations[i].push_back(seqNumber); // 2. sequence indices
152 /**************************************************************************************************/