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(), kmerSize(kSize) {
35 kmerDBName = fastaFileName.substr(0,fastaFileName.find_last_of(".")+1) + char('0'+ kmerSize) + "mer";
37 int power4s[14] = { 1, 4, 16, 64, 256, 1024, 4096, 16384, 65536, 262144, 1048576, 4194304, 16777216, 67108864 };
40 maxKmer = power4s[kmerSize];
41 kmerLocations.resize(maxKmer+1);
45 errorOut(e, "KmerDB", "KmerDB");
50 /**************************************************************************************************/
54 /**************************************************************************************************/
56 vector<int> KmerDB::findClosestSequences(Sequence* candidateSeq, int num){
58 vector<int> topMatches;
62 vector<int> matches(numSeqs, 0); // a record of the sequences with shared kmers
63 vector<int> timesKmerFound(kmerLocations.size()+1, 0); // a record of the kmers that we have already found
65 int numKmers = candidateSeq->getNumBases() - kmerSize + 1;
67 for(int i=0;i<numKmers;i++){
68 int kmerNumber = kmer.getKmerNumber(candidateSeq->getUnaligned(), i); // go through the query sequence and get a kmer number
69 if(timesKmerFound[kmerNumber] == 0){ // if we haven't seen it before...
70 for(int j=0;j<kmerLocations[kmerNumber].size();j++){//increase the count for each sequence that also has
71 matches[kmerLocations[kmerNumber][j]]++; // that kmer
74 timesKmerFound[kmerNumber] = 1; // ok, we've seen the kmer now
77 vector<seqMatch> seqMatches;
78 for(int i=0;i<numSeqs;i++){
79 seqMatch temp(i, matches[i]);
80 seqMatches.push_back(temp);
83 //sorts putting largest matches first
84 sort(seqMatches.begin(), seqMatches.end(), compareSeqMatches);
86 searchScore = seqMatches[0].match;
87 searchScore = 100 * searchScore / (float) numKmers; // return the Sequence object corresponding to the db
90 for (int i = 0; i < num; i++) {
91 topMatches.push_back(seqMatches[i].seq);
97 errorOut(e, "KmerDB", "findClosestSequences");
102 /**************************************************************************************************/
104 void KmerDB::generateDB(){
107 ofstream kmerFile; // once we have the kmerLocations folder print it out
108 openOutputFile(kmerDBName, kmerFile); // to a file
110 for(int i=0;i<maxKmer;i++){ // step through all of the possible kmer numbers
111 kmerFile << i << ' ' << kmerLocations[i].size(); // print the kmer number and the number of sequences with
112 for(int j=0;j<kmerLocations[i].size();j++){ // that kmer. then print out the indices of the sequences
113 kmerFile << ' ' << kmerLocations[i][j]; // with that kmer.
120 catch(exception& e) {
121 errorOut(e, "KmerDB", "generateDB");
126 /**************************************************************************************************/
127 void KmerDB::addSequence(Sequence seq) {
131 string unaligned = seq.getUnaligned(); // ...take the unaligned sequence...
132 int numKmers = unaligned.length() - kmerSize + 1;
134 vector<int> seenBefore(maxKmer+1,0);
135 for(int j=0;j<numKmers;j++){ // ...step though the sequence and get each kmer...
136 int kmerNumber = kmer.getKmerNumber(unaligned, j);
137 if(seenBefore[kmerNumber] == 0){
138 kmerLocations[kmerNumber].push_back(count); // ...insert the sequence index into kmerLocations for
139 } // the appropriate kmer number
140 seenBefore[kmerNumber] = 1;
145 catch(exception& e) {
146 errorOut(e, "KmerDB", "addSequence");
150 /**************************************************************************************************/
152 void KmerDB::readKmerDB(ifstream& kmerDBFile){
155 kmerDBFile.seekg(0); // start at the beginning of the file
160 for(int i=0;i<maxKmer;i++){
162 kmerDBFile >> seqName >> numValues;
164 for(int j=0;j<numValues;j++){ // for each kmer number get the...
165 kmerDBFile >> seqNumber; // 1. number of sequences with the kmer number
166 kmerLocations[i].push_back(seqNumber); // 2. sequence indices
172 catch(exception& e) {
173 errorOut(e, "KmerDB", "readKmerDB");
178 /**************************************************************************************************/
179 int KmerDB::getCount(int kmer) {
181 if (kmer < 0) { return 0; } //if user gives negative number
182 else if (kmer > maxKmer) { return 0; } //or a kmer that is bigger than maxkmer
183 else { return kmerLocations[kmer].size(); } // kmer is in vector range
185 catch(exception& e) {
186 errorOut(e, "KmerDB", "getCount");
190 /**************************************************************************************************/
191 vector<int> KmerDB::getSequencesWithKmer(int kmer) {
196 if (kmer < 0) { } //if user gives negative number
197 else if (kmer > maxKmer) { } //or a kmer that is bigger than maxkmer
198 else { seqs = kmerLocations[kmer]; }
202 catch(exception& e) {
203 errorOut(e, "KmerDB", "getSequencesWithKmer");
209 /**************************************************************************************************/