cluster.cpp

   1 /*
   2  *  cluster.cpp
   3  *
   4  *
   5  *  Created by Pat Schloss on 8/14/08.
   6  *  Copyright 2008 Patrick D. Schloss. All rights reserved.
   7  *
   8  */
   9
  10 #include "cluster.hpp"
  11 #include "rabundvector.hpp"
  12 #include "listvector.hpp"
  13 #include "sparsematrix.hpp"
  14
  15 /***********************************************************************/
  16
  17 Cluster::Cluster(RAbundVector* rav, ListVector* lv, SparseMatrix* dm, float c, string f) :
  18 rabund(rav), list(lv), dMatrix(dm), method(f)
  19 {
  20 /*
  21         cout << "sizeof(MatData): " << sizeof(MatData) << endl;
  22         cout << "sizeof(PCell*): " << sizeof(PCell*) << endl;
  23
  24         int nCells = dMatrix->getNNodes();
  25         time_t start = time(NULL);
  26
  27         MatVec matvec = MatVec(nCells);
  28         int i = 0;
  29         for (MatData currentCell = dMatrix->begin(); currentCell != dMatrix->end(); currentCell++) {
  30                 matvec[i++] = currentCell;
  31         }
  32         for (i= matvec.size();i>0;i--) {
  33                 dMatrix->rmCell(matvec[i-1]);
  34         }
  35         MatData it = dMatrix->begin();
  36         while (it != dMatrix->end()) {
  37                 it = dMatrix->rmCell(it);
  38         }
  39         cout << "Time to remove " << nCells << " cells: " << time(NULL) - start << " seconds" << endl;
  40     exit(0);
  41         MatData it = dMatrix->begin();
  42         cout << it->row << "/" << it->column << "/" << it->dist << endl;
  43         dMatrix->rmCell(dMatrix->begin());
  44         cout << it->row << "/" << it->column << "/" << it->dist << endl;
  45         exit(0);
  46 */
  47
  48         // Create a data structure to quickly access the PCell information
  49         // for a certain sequence. It consists of a vector of lists, where
  50         // a list contains pointers (iterators) to the all distances related
  51         // to a certain sequence. The Vector is accessed via the index of a
  52         // sequence in the distance matrix.
  53         seqVec = vector<MatVec>(lv->size());
  54         for (MatData currentCell = dMatrix->begin(); currentCell != dMatrix->end(); currentCell++) {
  55                 seqVec[currentCell->row].push_back(currentCell);
  56                 seqVec[currentCell->column].push_back(currentCell);
  57         }
  58         mapWanted = false;  //set to true by mgcluster to speed up overlap merge
  59
  60         //save so you can modify as it changes in average neighbor
  61         cutoff = c;
  62         m = MothurOut::getInstance();
  63 }
  64
  65 /***********************************************************************/
  66
  67 void Cluster::getRowColCells() {
  68         try {
  69                 PCell* smallCell = dMatrix->getSmallestCell();  //find the smallest cell - this routine should probably not be in the SpMat class
  70
  71                 smallRow = smallCell->row;              // get its row
  72                 smallCol = smallCell->column;   // get its column
  73                 smallDist = smallCell->dist;    // get the smallest distance
  74
  75                 rowCells = seqVec[smallRow];    // all distances related to the row index
  76                 colCells = seqVec[smallCol];    // all distances related to the column index
  77                 nRowCells = rowCells.size();
  78                 nColCells = colCells.size();
  79         }
  80         catch(exception& e) {
  81                 m->errorOut(e, "Cluster", "getRowColCells");
  82                 exit(1);
  83         }
  84
  85 }
  86 /***********************************************************************/
  87 // Remove the specified cell from the seqVec and from the sparse
  88 // matrix
  89 void Cluster::removeCell(const MatData& cell, int vrow, int vcol, bool rmMatrix)
  90 {
  91         ull drow = cell->row;
  92         ull dcol = cell->column;
  93         if (((vrow >=0) && (drow != smallRow)) ||
  94                 ((vcol >=0) && (dcol != smallCol))) {
  95                 ull dtemp = drow;
  96                 drow = dcol;
  97                 dcol = dtemp;
  98         }
  99
 100         ull crow;
 101         ull ccol;
 102         int nCells;
 103         if (vrow < 0) {
 104                 nCells = seqVec[drow].size();
 105                 for (vrow=0; vrow<nCells;vrow++) {
 106                         crow = seqVec[drow][vrow]->row;
 107                         ccol = seqVec[drow][vrow]->column;
 108                         if (((crow == drow) && (ccol == dcol)) ||
 109                                 ((ccol == drow) && (crow == dcol))) {
 110                                 break;
 111                         }
 112                 }
 113         }
 114         seqVec[drow].erase(seqVec[drow].begin()+vrow);
 115         if (vcol < 0) {
 116                 nCells = seqVec[dcol].size();
 117                 for (vcol=0; vcol<nCells;vcol++) {
 118                         crow = seqVec[dcol][vcol]->row;
 119                         ccol = seqVec[dcol][vcol]->column;
 120                         if (((crow == drow) && (ccol == dcol)) ||
 121                                 ((ccol == drow) && (crow == dcol))) {
 122                                 break;
 123                         }
 124                 }
 125         }
 126         seqVec[dcol].erase(seqVec[dcol].begin()+vcol);
 127         if (rmMatrix) {
 128                 dMatrix->rmCell(cell);
 129         }
 130 }
 131
 132
 133 /***********************************************************************/
 134
 135 void Cluster::clusterBins(){
 136         try {
 137         //      cout << smallCol << '\t' << smallRow << '\t' << smallDist << '\t' << rabund->get(smallRow) << '\t' << rabund->get(smallCol);
 138
 139                 rabund->set(smallCol, rabund->get(smallRow)+rabund->get(smallCol));
 140                 rabund->set(smallRow, 0);
 141                 rabund->setLabel(toString(smallDist));
 142
 143         //      cout << '\t' << rabund->get(smallRow) << '\t' << rabund->get(smallCol) << endl;
 144         }
 145         catch(exception& e) {
 146                 m->errorOut(e, "Cluster", "clusterBins");
 147                 exit(1);
 148         }
 149
 150
 151 }
 152
 153 /***********************************************************************/
 154
 155 void Cluster::clusterNames(){
 156         try {
 157         //      cout << smallCol << '\t' << smallRow << '\t' << smallDist << '\t' << list->get(smallRow) << '\t' << list->get(smallCol);
 158                 if (mapWanted) {  updateMap();  }
 159
 160                 list->set(smallCol, list->get(smallRow)+','+list->get(smallCol));
 161                 list->set(smallRow, "");
 162                 list->setLabel(toString(smallDist));
 163
 164         //      cout << '\t' << list->get(smallRow) << '\t' << list->get(smallCol) << endl;
 165     }
 166         catch(exception& e) {
 167                 m->errorOut(e, "Cluster", "clusterNames");
 168                 exit(1);
 169         }
 170
 171 }
 172
 173 /***********************************************************************/
 174 //This function clusters based on the method of the derived class
 175 //At the moment only average and complete linkage are covered, because
 176 //single linkage uses a different approach.
 177 void Cluster::update(double& cutOFF){
 178         try {
 179                 getRowColCells();
 180
 181                 vector<int> foundCol(nColCells, 0);
 182
 183                 int search;
 184                 bool changed;
 185
 186                 // The vector has to be traversed in reverse order to preserve the index
 187                 // for faster removal in removeCell()
 188                 for (int i=nRowCells-1;i>=0;i--) {
 189                         if (!((rowCells[i]->row == smallRow) && (rowCells[i]->column == smallCol))) {
 190                                 if (rowCells[i]->row == smallRow) {
 191                                         search = rowCells[i]->column;
 192                                 } else {
 193                                         search = rowCells[i]->row;
 194                                 }
 195
 196                                 bool merged = false;
 197                                 for (int j=0;j<nColCells;j++) {
 198                                         if (!((colCells[j]->row == smallRow) && (colCells[j]->column == smallCol))) { //if you are not hte smallest distance
 199                                                 if (colCells[j]->row == search || colCells[j]->column == search) {
 200                                                         foundCol[j] = 1;
 201                                                         merged = true;
 202                                                         changed = updateDistance(colCells[j], rowCells[i]);
 203                                                         // If the cell's distance changed and it had the same distance as
 204                                                         // the smallest distance, invalidate the mins vector in SparseMatrix
 205                                                         if (changed) {
 206                                                                 if (colCells[j]->vectorMap != NULL) {
 207                                                                         *(colCells[j]->vectorMap) = NULL;
 208                                                                         colCells[j]->vectorMap = NULL;
 209                                                                 }
 210                                                         }
 211                                                         break;
 212                                                 }
 213                                         }
 214                                 }
 215                                 //if not merged it you need it for warning
 216                                 if ((!merged) && (method == "average")) {
 217                                         //m->mothurOut("Warning: trying to merge cell " + toString(rowCells[i]->row+1) + " " + toString(rowCells[i]->column+1) + " distance " + toString(rowCells[i]->dist) + " with value above cutoff. Results may vary from using cutoff at cluster command instead of read.dist."); m->mothurOutEndLine();
 218                                         if (cutOFF > rowCells[i]->dist) {
 219                                                 cutOFF = rowCells[i]->dist;
 220                                                 //m->mothurOut("changing cutoff to " + toString(cutOFF));  m->mothurOutEndLine();
 221                                         }
 222
 223                                 }
 224                                 removeCell(rowCells[i], i , -1);
 225
 226                         }
 227                 }
 228                 clusterBins();
 229                 clusterNames();
 230
 231                 // Special handling for singlelinkage case, not sure whether this
 232                 // could be avoided
 233                 for (int i=nColCells-1;i>=0;i--) {
 234                         if (foundCol[i] == 0) {
 235                                 if (method == "average") {
 236                                         if (!((colCells[i]->row == smallRow) && (colCells[i]->column == smallCol))) {
 237                                                 //m->mothurOut("Warning: merging cell " + toString(colCells[i]->row+1) + " " + toString(colCells[i]->column+1) + " distance " + toString(colCells[i]->dist) + " value above cutoff. Results may vary from using cutoff at cluster command instead of read.dist."); m->mothurOutEndLine();
 238                                                 if (cutOFF > colCells[i]->dist) {
 239                                                         cutOFF = colCells[i]->dist;
 240                                                         //m->mothurOut("changing cutoff to " + toString(cutOFF));  m->mothurOutEndLine();
 241                                                 }
 242                                         }
 243                                 }
 244                                 removeCell(colCells[i], -1, i);
 245                         }
 246                 }
 247         }
 248         catch(exception& e) {
 249                 m->errorOut(e, "Cluster", "update");
 250                 exit(1);
 251         }
 252 }
 253 /***********************************************************************/
 254 void Cluster::setMapWanted(bool f)  {
 255         try {
 256                 mapWanted = f;
 257
 258                 //initialize map
 259                 for (int i = 0; i < list->getNumBins(); i++) {
 260
 261                         //parse bin
 262                         string names = list->get(i);
 263                         while (names.find_first_of(',') != -1) {
 264                                 //get name from bin
 265                                 string name = names.substr(0,names.find_first_of(','));
 266                                 //save name and bin number
 267                                 seq2Bin[name] = i;
 268                                 names = names.substr(names.find_first_of(',')+1, names.length());
 269                         }
 270
 271                         //get last name
 272                         seq2Bin[names] = i;
 273                 }
 274
 275         }
 276         catch(exception& e) {
 277                 m->errorOut(e, "Cluster", "setMapWanted");
 278                 exit(1);
 279         }
 280 }
 281 /***********************************************************************/
 282 void Cluster::updateMap() {
 283 try {
 284                 //update location of seqs in smallRow since they move to smallCol now
 285                 string names = list->get(smallRow);
 286                 while (names.find_first_of(',') != -1) {
 287                         //get name from bin
 288                         string name = names.substr(0,names.find_first_of(','));
 289                         //save name and bin number
 290                         seq2Bin[name] = smallCol;
 291                         names = names.substr(names.find_first_of(',')+1, names.length());
 292                 }
 293
 294                 //get last name
 295                 seq2Bin[names] = smallCol;
 296
 297         }
 298         catch(exception& e) {
 299                 m->errorOut(e, "Cluster", "updateMap");
 300                 exit(1);
 301         }
 302 }
 303 /***********************************************************************/
 304
 305
 306