added cluster.split command

[mothur.git] / blastdb.cpp

diff --git a/blastdb.cpp b/blastdb.cpp
index 6b45a48ce379168e82300209a7668c3e07e4d0ab..35a19f485816c87db85b71c74ca138f00595b180 100644 (file)
--- a/blastdb.cpp
+++ b/blastdb.cpp
@@ -25,6 +25,19 @@ gapOpen(gO), gapExtend(gE), match(m), misMatch(mM) {
        queryFileName = toString(randNumber) + ".candidate.unaligned.fasta";
        blastFileName = toString(randNumber) + ".blast";
 
+}
+/**************************************************************************************************/
+
+BlastDB::BlastDB() : Database() {
+       
+       globaldata = GlobalData::getInstance();
+       count = 0;
+
+       int randNumber = rand();
+       dbFileName = toString(randNumber) + ".template.unaligned.fasta";
+       queryFileName = toString(randNumber) + ".candidate.unaligned.fasta";
+       blastFileName = toString(randNumber) + ".blast";
+
 }
 
 /**************************************************************************************************/
@@ -50,7 +63,7 @@ vector<int> BlastDB::findClosestSequences(Sequence* seq, int n) {
                //      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/blastall -p blastn -d " + dbFileName + " -b 1 -m 8 -W 28 -v " + toString(n);
+               string blastCommand = path + "blast/bin/blastall -p blastn -d " + dbFileName + " -m 8 -W 28 -v " + toString(n) + " -b " + toString(n);;
                blastCommand += (" -i " + queryFileName + " -o " + blastFileName);
                system(blastCommand.c_str());
                
@@ -75,12 +88,58 @@ vector<int> BlastDB::findClosestSequences(Sequence* seq, int n) {
                return topMatches;
        }
        catch(exception& e) {
-               errorOut(e, "BlastDB", "findClosestSequences");
+               m->errorOut(e, "BlastDB", "findClosestSequences");
                exit(1);
        }
 
 }
 /**************************************************************************************************/
+//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);
+       }
+}
+/**************************************************************************************************/
 void BlastDB::addSequence(Sequence seq) {
        try {
        
@@ -91,11 +150,11 @@ void BlastDB::addSequence(Sequence seq) {
                unalignedFastaFile << '>' << count << endl;                                     //      sequences, which will be input to formatdb
                unalignedFastaFile << seq.getUnaligned() << endl;
                unalignedFastaFile.close();
-               
+       
                count++;
        }
        catch(exception& e) {
-               errorOut(e, "BlastDB", "addSequence");
+               m->errorOut(e, "BlastDB", "addSequence");
                exit(1);
        }
 }
@@ -103,7 +162,7 @@ void BlastDB::addSequence(Sequence seq) {
 void BlastDB::generateDB() {
        try {
        
-               mothurOut("Generating the temporary BLAST database...\t");      cout.flush();
+               //m->mothurOut("Generating the temporary BLAST database...\t"); cout.flush();
                
                path = globaldata->argv;
                path = path.substr(0, (path.find_last_of('m')));
@@ -111,13 +170,14 @@ void BlastDB::generateDB() {
                string formatdbCommand = path + "blast/bin/formatdb -p F -o T -i " + dbFileName;        //      format the database, -o option gives us the ability
                system(formatdbCommand.c_str());                                                                //      to get the right sequence names, i think. -p F
                                                                                                                                        //      option tells formatdb that seqs are DNA, not prot
-               mothurOut("DONE."); mothurOutEndLine(); mothurOutEndLine(); cout.flush();
+               //m->mothurOut("DONE."); m->mothurOutEndLine(); m->mothurOutEndLine(); cout.flush();
        }
        catch(exception& e) {
-               errorOut(e, "BlastDB", "generateDB");
+               m->errorOut(e, "BlastDB", "generateDB");
                exit(1);
        }
 }
+/**************************************************************************************************/
 
 /**************************************************************************************************/