+//assumes you have added all the template sequences using the addSequence function and run generateDB.
+vector<int> BlastDB::findClosestMegaBlast(Sequence* seq, int n) {
+ try{
+ vector<int> topMatches;
+
+ ofstream queryFile;
+ openOutputFile(queryFileName, queryFile);
+ queryFile << '>' << seq->getName() << endl;
+ queryFile << seq->getUnaligned() << endl;
+ queryFile.close();
+
+ // the goal here is to quickly survey the database to find the closest match. To do this we are using the default
+ // wordsize used in megablast. I'm sure we're sacrificing accuracy for speed, but anyother way would take way too
+ // long. With this setting, it seems comparable in speed to the suffix tree approach.
+
+ string blastCommand = path + "blast/bin/megablast -e 1e-10 -d " + dbFileName + " -m 8 -b " + toString(n) + " -v " + toString(n); //-W 28 -p blastn
+ blastCommand += (" -i " + queryFileName + " -o " + blastFileName);
+ system(blastCommand.c_str());
+
+ ifstream m8FileHandle;
+ openInputFile(blastFileName, m8FileHandle, "no error");
+
+ string dummy;
+ int templateAccession;
+ gobble(m8FileHandle);
+
+ while(!m8FileHandle.eof()){
+ m8FileHandle >> dummy >> templateAccession >> searchScore;
+
+ //get rest of junk in line
+ while (!m8FileHandle.eof()) { char c = m8FileHandle.get(); if (c == 10 || c == 13){ break; } }
+
+ gobble(m8FileHandle);
+ topMatches.push_back(templateAccession);
+//cout << templateAccession << endl;
+ }
+ m8FileHandle.close();
+//cout << "\n\n" ;
+ return topMatches;
+ }
+ catch(exception& e) {
+ m->errorOut(e, "BlastDB", "findClosest");
+ exit(1);
+ }
+}
+/**************************************************************************************************/