5 * Created by westcott on 1/11/11.
6 * Copyright 2011 Schloss Lab. All rights reserved.
10 #include "nmdscommand.h"
11 #include "readphylipvector.h"
13 //**********************************************************************************************************************
14 vector<string> NMDSCommand::getValidParameters(){
16 string Array[] = {"phylip","axes","mindim","maxdim","iters","maxiters","epsilon","outputdir","inputdir"};
17 vector<string> myArray (Array, Array+(sizeof(Array)/sizeof(string)));
21 m->errorOut(e, "NMDSCommand", "getValidParameters");
25 //**********************************************************************************************************************
26 NMDSCommand::NMDSCommand(){
28 abort = true; calledHelp = true;
29 vector<string> tempOutNames;
30 outputTypes["nmds"] = tempOutNames;
31 outputTypes["stress"] = tempOutNames;
32 outputTypes["iters"] = tempOutNames;
35 m->errorOut(e, "NMDSCommand", "NMDSCommand");
39 //**********************************************************************************************************************
40 vector<string> NMDSCommand::getRequiredParameters(){
42 string Array[] = {"phylip"};
43 vector<string> myArray (Array, Array+(sizeof(Array)/sizeof(string)));
47 m->errorOut(e, "NMDSCommand", "getRequiredParameters");
51 //**********************************************************************************************************************
52 vector<string> NMDSCommand::getRequiredFiles(){
54 vector<string> myArray;
58 m->errorOut(e, "NMDSCommand", "getRequiredFiles");
62 //**********************************************************************************************************************
64 NMDSCommand::NMDSCommand(string option) {
66 abort = false; calledHelp = false;
68 //allow user to run help
69 if(option == "help") { help(); abort = true; calledHelp = true; }
72 //valid paramters for this command
73 string Array[] = {"phylip","axes","mindim","maxdim","iters","maxiters","epsilon","outputdir", "inputdir"};
74 vector<string> myArray (Array, Array+(sizeof(Array)/sizeof(string)));
76 OptionParser parser(option);
77 map<string, string> parameters = parser. getParameters();
79 ValidParameters validParameter;
80 map<string, string>::iterator it;
82 //check to make sure all parameters are valid for command
83 for (it = parameters.begin(); it != parameters.end(); it++) {
84 if (validParameter.isValidParameter(it->first, myArray, it->second) != true) { abort = true; }
86 //if the user changes the input directory command factory will send this info to us in the output parameter
87 string inputDir = validParameter.validFile(parameters, "inputdir", false);
88 if (inputDir == "not found"){ inputDir = ""; }
91 it = parameters.find("phylip");
92 //user has given a template file
93 if(it != parameters.end()){
94 path = m->hasPath(it->second);
95 //if the user has not given a path then, add inputdir. else leave path alone.
96 if (path == "") { parameters["phylip"] = inputDir + it->second; }
99 it = parameters.find("axes");
100 //user has given a template file
101 if(it != parameters.end()){
102 path = m->hasPath(it->second);
103 //if the user has not given a path then, add inputdir. else leave path alone.
104 if (path == "") { parameters["axes"] = inputDir + it->second; }
108 //initialize outputTypes
109 vector<string> tempOutNames;
110 outputTypes["nmds"] = tempOutNames;
111 outputTypes["iters"] = tempOutNames;
112 outputTypes["stress"] = tempOutNames;
114 //required parameters
115 phylipfile = validParameter.validFile(parameters, "phylip", true);
116 if (phylipfile == "not open") { phylipfile = ""; abort = true; }
117 else if (phylipfile == "not found") { phylipfile = ""; m->mothurOut("You must provide a distance file before running the nmds command."); m->mothurOutEndLine(); abort = true; }
119 axesfile = validParameter.validFile(parameters, "axes", true);
120 if (axesfile == "not open") { axesfile = ""; abort = true; }
121 else if (axesfile == "not found") { axesfile = ""; }
123 //if the user changes the output directory command factory will send this info to us in the output parameter
124 outputDir = validParameter.validFile(parameters, "outputdir", false); if (outputDir == "not found"){
126 outputDir += m->hasPath(phylipfile); //if user entered a file with a path then preserve it
129 string temp = validParameter.validFile(parameters, "mindim", false); if (temp == "not found") { temp = "2"; }
130 convert(temp, mindim);
132 temp = validParameter.validFile(parameters, "maxiters", false); if (temp == "not found") { temp = "500"; }
133 convert(temp, maxIters);
135 temp = validParameter.validFile(parameters, "iters", false); if (temp == "not found") { temp = "10"; }
136 convert(temp, iters);
138 temp = validParameter.validFile(parameters, "maxdim", false); if (temp == "not found") { temp = "2"; }
139 convert(temp, maxdim);
141 temp = validParameter.validFile(parameters, "epsilon", false); if (temp == "not found") { temp = "0.000000000001"; }
142 convert(temp, epsilon);
144 if (mindim < 1) { m->mothurOut("mindim must be at least 1."); m->mothurOutEndLine(); abort = true; }
145 if (maxdim < mindim) { m->mothurOut("maxdim must be greater than mindim."); m->mothurOutEndLine(); abort = true; }
149 catch(exception& e) {
150 m->errorOut(e, "NMDSCommand", "NMDSCommand");
154 //**********************************************************************************************************************
155 void NMDSCommand::help(){
157 m->mothurOut("The nmds command is modelled after the nmds code written in R by Sarah Goslee, using Non-metric multidimensional scaling function using the majorization algorithm from Borg & Groenen 1997, Modern Multidimensional Scaling."); m->mothurOutEndLine();
158 m->mothurOut("The nmds command parameters are phylip, axes, mindim, maxdim, maxiters, iters and epsilon."); m->mothurOutEndLine();
159 m->mothurOut("The phylip parameter allows you to enter your distance file."); m->mothurOutEndLine();
160 m->mothurOut("The axes parameter allows you to enter a file containing a starting configuration."); m->mothurOutEndLine();
161 m->mothurOut("The maxdim parameter allows you to select the maximum dimensions to use. Default=2"); m->mothurOutEndLine();
162 m->mothurOut("The mindim parameter allows you to select the minimum dimensions to use. Default=2"); m->mothurOutEndLine();
163 m->mothurOut("The maxiters parameter allows you to select the maximum number of iters to try with each random configuration. Default=500"); m->mothurOutEndLine();
164 m->mothurOut("The iters parameter allows you to select the number of random configuration to try. Default=10"); m->mothurOutEndLine();
165 m->mothurOut("The epsilon parameter allows you to select set an acceptable stopping point. Default=1e-12."); m->mothurOutEndLine();
166 m->mothurOut("Example nmds(phylip=yourDistanceFile).\n");
167 m->mothurOut("Note: No spaces between parameter labels (i.e. phylip), '=' and parameters (i.e.yourDistanceFile).\n\n");
169 catch(exception& e) {
170 m->errorOut(e, "NMDSCommand", "help");
174 //**********************************************************************************************************************
175 NMDSCommand::~NMDSCommand(){}
176 //**********************************************************************************************************************
177 int NMDSCommand::execute(){
180 if (abort == true) { if (calledHelp) { return 0; } return 2; }
182 cout.setf(ios::fixed, ios::floatfield);
183 cout.setf(ios::showpoint);
185 vector<string> names;
186 vector< vector< double> > matrix;
188 //read in phylip file
189 ReadPhylipVector readFile(phylipfile);
190 names = readFile.read(matrix);
191 if (m->control_pressed) { return 0; }
194 vector< vector<double> > axes;
195 if (axesfile != "") { axes = readAxes(names); }
197 string outputFileName = outputDir + m->getRootName(m->getSimpleName(phylipfile)) + "nmds.iters";
198 string stressFileName = outputDir + m->getRootName(m->getSimpleName(phylipfile)) + "nmds.stress";
199 outputNames.push_back(outputFileName); outputTypes["iters"].push_back(outputFileName);
200 outputNames.push_back(stressFileName); outputTypes["stress"].push_back(stressFileName);
203 m->openOutputFile(outputFileName, out);
204 m->openOutputFile(stressFileName, out2);
206 out2.setf(ios::fixed, ios::floatfield);
207 out2.setf(ios::showpoint);
208 out.setf(ios::fixed, ios::floatfield);
209 out.setf(ios::showpoint);
211 out2 << "Dimension\tIter\tStress\tRsq" << endl;
213 double bestStress = 10000000;
214 double bestR2 = 10000000;
215 vector< vector<double> > bestConfig;
218 for (int i = mindim; i <= maxdim; i++) {
219 m->mothurOut("Processing Dimension: " + toString(i)); m->mothurOutEndLine();
221 for (int j = 0; j < iters; j++) {
222 m->mothurOut(toString(j+1)); m->mothurOutEndLine();
224 //get configuration - either randomly generate or resize to this dimension
225 vector< vector<double> > thisConfig;
226 if (axesfile == "") { thisConfig = generateStartingConfiguration(names.size(), i); }
227 else { thisConfig = getConfiguration(axes, i); }
228 if (m->control_pressed) { out.close(); out2.close(); for (int k = 0; k < outputNames.size(); k++) { remove(outputNames[k].c_str()); } return 0; }
230 //calc nmds for this dimension
232 vector< vector<double> > endConfig = nmdsCalc(matrix, thisConfig, stress);
233 if (m->control_pressed) { out.close(); out2.close(); for (int k = 0; k < outputNames.size(); k++) { remove(outputNames[k].c_str()); } return 0; }
235 //calc euclid distances for new config
236 vector< vector<double> > newEuclid = linearCalc.calculateEuclidianDistance(endConfig);
237 if (m->control_pressed) { out.close(); out2.close(); for (int k = 0; k < outputNames.size(); k++) { remove(outputNames[k].c_str()); } return 0; }
239 //calc correlation between original distances and euclidean distances from this config
240 double rsquared = linearCalc.calcPearson(newEuclid, matrix);
241 rsquared *= rsquared;
242 if (m->control_pressed) { out.close(); out2.close(); for (int k = 0; k < outputNames.size(); k++) { remove(outputNames[k].c_str()); } return 0; }
245 out << "Config" << (j+1) << '\t';
246 for (int k = 0; k < i; k++) { out << "axis" << (k+1) << '\t'; }
248 out2 << i << '\t' << (j+1) << '\t' << stress << '\t' << rsquared << endl;
250 output(endConfig, names, out);
253 if (stress < bestStress) {
257 bestConfig = endConfig;
260 if (m->control_pressed) { out.close(); out2.close(); for (int k = 0; k < outputNames.size(); k++) { remove(outputNames[k].c_str()); } return 0; }
264 out.close(); out2.close();
267 string BestFileName = outputDir + m->getRootName(m->getSimpleName(phylipfile)) + "nmds.axes";
268 outputNames.push_back(BestFileName); outputTypes["nmds"].push_back(BestFileName);
270 m->mothurOut("\nNumber of dimensions:\t" + toString(bestDim) + "\n");
271 m->mothurOut("Lowest stress :\t" + toString(bestStress) + "\n");
272 m->mothurOut("R-squared for configuration:\t" + toString(bestR2) + "\n");
275 m->openOutputFile(BestFileName, outBest);
276 outBest.setf(ios::fixed, ios::floatfield);
277 outBest.setf(ios::showpoint);
280 for (int k = 0; k < bestConfig.size(); k++) { outBest << "axis" << (k+1) << '\t'; }
283 output(bestConfig, names, outBest);
287 if (m->control_pressed) { for (int i = 0; i < outputNames.size(); i++) { remove(outputNames[i].c_str()); } return 0; }
289 m->mothurOutEndLine();
290 m->mothurOut("Output File Names: "); m->mothurOutEndLine();
291 for (int i = 0; i < outputNames.size(); i++) { m->mothurOut(outputNames[i]); m->mothurOutEndLine(); }
292 m->mothurOutEndLine();
296 catch(exception& e) {
297 m->errorOut(e, "NMDSCommand", "execute");
301 //**********************************************************************************************************************
302 vector< vector<double> > NMDSCommand::nmdsCalc(vector< vector<double> >& matrix, vector< vector<double> >& config, double& stress1) {
305 vector< vector<double> > newConfig = config;
307 //calc euclid distances
308 vector< vector<double> > euclid = linearCalc.calculateEuclidianDistance(newConfig);
309 if (m->control_pressed) { return newConfig; }
311 double stress2 = calculateStress(matrix, euclid);
312 stress1 = stress2 + 1.0 + epsilon;
315 while ((count < maxIters) && (abs(stress1 - stress2) > epsilon)) {
320 if (m->control_pressed) { return newConfig; }
322 vector< vector<double> > b; b.resize(euclid.size());
323 for (int i = 0; i < b.size(); i++) { b[i].resize(euclid[i].size(), 0.0); }
325 vector<double> columnSums; columnSums.resize(euclid.size(), 0.0);
326 for (int i = 0; i < euclid.size(); i++) {
327 for (int j = 0; j < euclid[i].size(); j++) {
328 //eliminate divide by zero error
329 if (euclid[i][j] != 0) {
330 b[i][j] = matrix[i][j] / euclid[i][j];
331 columnSums[j] += b[i][j];
337 //put in diagonal sums
338 for (int i = 0; i < euclid.size(); i++) { b[i][i] = columnSums[i]; }
340 int numInLowerTriangle = matrix.size() * (matrix.size()-1) / 2.0;
341 double n = (1.0 + sqrt(1.0 + 8.0 * numInLowerTriangle)) / 2.0;
344 newConfig = linearCalc.matrix_mult(newConfig, b);
345 for (int i = 0; i < newConfig.size(); i++) {
346 for (int j = 0; j < newConfig[i].size(); j++) {
347 newConfig[i][j] *= (1.0 / n);
351 euclid = linearCalc.calculateEuclidianDistance(newConfig);
353 stress2 = calculateStress(matrix, euclid);
358 catch(exception& e) {
359 m->errorOut(e, "NMDSCommand", "generateStartingConfiguration");
364 //**********************************************************************************************************************
365 //generate random config
366 vector< vector<double> > NMDSCommand::generateStartingConfiguration(int numNames, int dimension) {
368 vector< vector<double> > axes; axes.resize(dimension);
369 for (int i = 0; i < axes.size(); i++) { axes[i].resize(numNames); }
371 //generate random number between -1 and 1, precision 6
372 for (int i = 0; i < axes.size(); i++) {
373 for (int j = 0; j < axes[i].size(); j++) {
375 if (m->control_pressed) { return axes; }
377 //generate random int between 0 and 99999
378 int myrand = (int)((float)(rand()) / ((RAND_MAX / 99998) + 1));
380 //generate random sign
381 int mysign = (int)((float)(rand()) / ((RAND_MAX / 99998) + 1));
383 //if mysign is even then sign = positive, else sign = negative
384 if ((mysign % 2) == 0) { mysign = 1.0; }
385 else { mysign = -1.0; }
387 axes[i][j] = mysign * myrand / (float) 100000;
393 catch(exception& e) {
394 m->errorOut(e, "NMDSCommand", "generateStartingConfiguration");
398 //**********************************************************************************************************************
399 //normalize configuration
400 int NMDSCommand::normalizeConfiguration(vector< vector<double> >& axes, int numNames, int dimension) {
402 vector<double> averageAxes; averageAxes.resize(dimension, 0.0);
405 for (int i = 0; i < axes.size(); i++) {
406 for (int j = 0; j < axes[i].size(); j++) { averageAxes[i] += axes[i][j]; }
408 averageAxes[i] /= (float) numNames;
412 double sumDenom = 0.0;
413 for (int i = 0; i < axes.size(); i++) {
414 for (int j = 0; j < axes[i].size(); j++) {
415 sumDenom += ((axes[i][j] - averageAxes[i]) * (axes[i][j] - averageAxes[i]));
419 double denom = sqrt((sumDenom / (float) (axes.size() * numNames)));
421 for (int i = 0; i < axes.size(); i++) {
422 for (int j = 0; j < axes[i].size(); j++) {
423 axes[i][j] = (axes[i][j] - averageAxes[i]) / denom;
429 catch(exception& e) {
430 m->errorOut(e, "NMDSCommand", "normalizeConfiguration");
434 //**********************************************************************************************************************
436 vector< vector<double> > NMDSCommand::getConfiguration(vector< vector<double> >& axes, int dimension) {
438 vector< vector<double> > newAxes; newAxes.resize(dimension);
440 for (int i = 0; i < dimension; i++) {
441 newAxes[i] = axes[i];
446 catch(exception& e) {
447 m->errorOut(e, "NMDSCommand", "getConfiguration");
451 //**********************************************************************************************************************
452 //find raw stress, and normalize using
453 double NMDSCommand::calculateStress(vector< vector<double> >& matrix, vector< vector<double> >& config) {
455 double normStress = 0.0;
457 double rawStress = 0.0;
460 for (int i = 0; i < matrix.size(); i++) {
461 for (int j = 0; j < matrix[i].size(); j++) {
462 if (m->control_pressed) { return normStress; }
464 rawStress += ((matrix[i][j] - config[i][j]) * (matrix[i][j] - config[i][j]));
465 denom += (config[i][j] * config[i][j]);
470 if ((rawStress != 0.0) && (denom != 0.0)) {
471 normStress = sqrt((rawStress / denom));
476 catch(exception& e) {
477 m->errorOut(e, "NMDSCommand", "calculateStress");
482 //**********************************************************************************************************************
483 int NMDSCommand::output(vector< vector<double> >& config, vector<string>& names, ofstream& out) {
486 for (int i = 0; i < names.size(); i++) {
488 out << names[i] << '\t';
490 for (int j = 0; j < config.size(); j++) {
491 out << config[j][i] << '\t';
501 catch(exception& e) {
502 m->errorOut(e, "NMDSCommand", "output");
506 /*****************************************************************/
507 vector< vector<double> > NMDSCommand::readAxes(vector<string> names){
510 m->openInputFile(axesfile, in);
512 string headerLine = m->getline(in); m->gobble(in);
514 //count the number of axis you are reading
518 int pos = headerLine.find("axis");
519 if (pos != string::npos) {
521 headerLine = headerLine.substr(pos+4);
522 }else { done = true; }
525 if (maxdim > count) {
526 m->mothurOut("You requested maxdim = " + toString(maxdim) + ", but your file only includes " + toString(count) + ". Using " + toString(count) + "."); m->mothurOutEndLine();
528 if (maxdim < mindim) { m->mothurOut("Also adjusting mindim to " + toString(maxdim-1) + "."); m->mothurOutEndLine(); }
531 vector< vector<double> > axes; axes.resize(maxdim);
532 for (int i = 0; i < axes.size(); i++) { axes[i].resize(names.size(), 0.0); }
534 map <string, vector<double> > orderedAxes;
535 map <string, vector<double> >::iterator it;
539 if (m->control_pressed) { in.close(); return axes; }
542 in >> group; m->gobble(in);
545 if (!m->inUsersGroups(group, names)) { ignore = true; m->mothurOut(group + " is in your axes file and not in your distance file, ignoring."); m->mothurOutEndLine(); }
547 vector<double> thisGroupsAxes;
548 for (int i = 0; i < count; i++) {
552 //only save the axis we want
553 if (i < maxdim) { thisGroupsAxes.push_back(temp); }
556 if (!ignore) { orderedAxes[group] = thisGroupsAxes; }
563 if (names.size() != orderedAxes.size()) { m->mothurOut("[ERROR]: your axes file does not match your distance file, aborting."); m->mothurOutEndLine(); m->control_pressed = true; return axes; }
565 //put axes info in same order as distance file, just in case
566 for (int i = 0; i < names.size(); i++) {
567 it = orderedAxes.find(names[i]);
569 if (it != orderedAxes.end()) {
570 vector<double> thisGroupsAxes = it->second;
572 for (int j = 0; j < thisGroupsAxes.size(); j++) {
573 axes[j][i] = thisGroupsAxes[j];
576 }else { m->mothurOut("[ERROR]: your axes file does not match your distance file, aborting."); m->mothurOutEndLine(); m->control_pressed = true; return axes; }
581 catch(exception& e) {
582 m->errorOut(e, "NMDSCommand", "readAxes");
586 /**********************************************************************************************************************/