public boolean cluster(String[] args) {
boolean success = false;
// handle command line arguments; sets dataFile and threshold
success = parseArgs(args);
Map<String, Map<Integer, List<Isolate>>> dataMap =
new HashMap<String, Map<Integer, List<Isolate>>>();
/*
* Each isolate similarity matrix holds a correlation for both regions
*/
IsolateSimilarityMatrix partialCorrelations = new IsolateSimilarityMatrix();
Map<Connectivity, IsolateSimilarityMatrix> isolateNetworks =
new HashMap<Connectivity, IsolateSimilarityMatrix>();
isolateNetworks.put(Connectivity.STRONG, new IsolateSimilarityMatrix());
isolateNetworks.put(Connectivity.WEAK, new IsolateSimilarityMatrix());
Cluster.distType type = null;
String dataFileName = null;
double lowerThreshold = -1, upperThreshold = -1;
for (File dataFile : dataFileMap.keySet()) {
FileSettings settings = dataFileMap.get(dataFile);
dataFileName = dataFile.getName();
IsolateRegion region = settings.getRegion();
type = settings.getDistanceType();
lowerThreshold = settings.getLowerThreshold();
upperThreshold = settings.getUpperThreshold();
if (dataFile != null) {
IsolateFileParser parser = new IsolateFileParser(dataFile, settings);
// MARKER old code
// isolateMap = parser.extractData(similarityMatrix);
// mIsolateNetworks = parser.extractData();
parser.extractData(isolateNetworks, partialCorrelations);
}
System.out.println("strong Network size: " + isolateNetworks.get(Connectivity.STRONG).size());
}
// each point is a cluster, and we will combine two clusters in each iteration
List<ClusterDendogram> clustDends = clusterIsolates(type, isolateNetworks);
// System.err.println("clustDends length: " + clustDends.size());
// if the isolates yielded NO clusters (wtf data disappear?) or
// if clusterIsolates returned null, then this was *NOT* a success
success = clustDends != null && !clustDends.isEmpty();
String outputFileDir =
String.format("ClusterResults/%s_%.02f_%.02f", type, lowerThreshold, upperThreshold);
/*
String outputFileName = String.format("%s/%s", outputFileDir,
dataFile.getName().substring(0,
dataFile.getName().indexOf(".csv")));
*/
String origFileName =
HClustering.mOutputFileName.equals("")
? dataFileName.substring(0, dataFileName.indexOf(".csv"))
: HClustering.mOutputFileName;
System.out.println(
"Writing to file "
+ origFileName
+ " even though outputFilename should be "
+ HClustering.mOutputFileName);
String outputFileName = String.format("%s/%s", outputFileDir, origFileName);
IsolateOutputWriter.outputClusters(clustDends, outputFileDir, outputFileName + ".xml");
IsolateOutputWriter.outputCytoscapeFormat(clustDends, outputFileName);
IsolateOutputWriter.outputTemporalClusters(clustDends, outputFileName);
IsolateOutputWriter.outputTemporalCharts(clustDends, outputFileName);
return success;
}
private List<ClusterDendogram> clusterIsolates(
Cluster.distType type, Map<Connectivity, IsolateSimilarityMatrix> isolateNetworks) {
/*
IsolateRegion region = settings.getRegion();
double distanceThreshold = settings.getDistanceThreshold();
double lowerThreshold = settings.getLowerThreshold();
double upperThreshold = settings.getUpperThreshold();
Cluster.distType type = settings.getDistanceType();
*/
// mappings represent days to isolates
Map<String, Map<Integer, List<Isolate>>> technicianIsolateMap = null;
// list of all constructed clusters
List<ClusterDendogram> clusters = new ArrayList<ClusterDendogram>();
List<ClusterDendogram> technicianClusters = new ArrayList<ClusterDendogram>();
/*
* Marker
*/
// MARKER new code. get the isolateMap from the similarity matrix now
// also, multiple similarity matrices are stored in the isolate networks
// map so that it is possible to iterate on correlations based on their
// strength instead of just going all willy nilly
IsolateSimilarityMatrix similarityMatrix = isolateNetworks.get(Connectivity.STRONG);
technicianIsolateMap = similarityMatrix.getIsolateMap();
for (String technician : technicianIsolateMap.keySet()) {
System.out.printf("clustering technician %s's dataset...\n", technician);
Map<Integer, List<Isolate>> isolateMap = technicianIsolateMap.get(technician);
for (int sampleDay : isolateMap.keySet()) {
System.out.printf("clustering day %d...\n", sampleDay);
// Cluster the list of isolates in this day
List<ClusterDendogram> currClusters =
clusterIsolateList(similarityMatrix, isolateMap.get(sampleDay), type);
IsolateOutputWriter.outputClustersByDay(similarityMatrix, sampleDay, currClusters);
/*
System.err.println("currClusters length: " + currClusters.size());
/*
for (ClusterDendogram clustDend : currClusters) {
System.out.println(clustDend.getDendogram().getXML());
}
*/
// System.err.printf("on day %d there are a total of %d clusters", sampleDay,
// clusters.size());
// Cluster all previous days with this day
clusters = clusterToDate(clusters, currClusters, type);
}
technicianClusters.addAll(clusters);
clusters = new ArrayList<ClusterDendogram>();
}
clusters = clusterGroup(technicianClusters, type);
System.out.printf(
"\n\n================\nFINISHED STAGE 1. CLUSTERING SQUISHIES\n"
+ "===================\n\n");
similarityMatrix = isolateNetworks.get(Connectivity.WEAK);
technicianIsolateMap = similarityMatrix.getIsolateMap();
/*
* this is so that when doing the second pass through all of the clusters
* squishy correlations will also be known
*/
for (ClusterDendogram clust : clusters) {
clust.getCluster().setSimilarityMatrix(similarityMatrix);
}
for (String technician : technicianIsolateMap.keySet()) {
System.out.printf("clustering technician %s's dataset...\n", technician);
Map<Integer, List<Isolate>> isolateMap = technicianIsolateMap.get(technician);
for (int sampleDay : isolateMap.keySet()) {
System.out.printf("clustering day %d...\n", sampleDay);
for (Isolate isolate : isolateMap.get(sampleDay)) {
// clusters = clusterWeakIsolates(similarityMatrix, clusters, isolate, type);
if (!isolate.hasBeenClustered()) {
Cluster newCluster = new Cluster(similarityMatrix, isolate);
Dendogram newDendogram = new DendogramLeaf(isolate);
clusters.add(new ClusterDendogram(newCluster, newDendogram));
}
}
}
}
clusters = clusterGroup(clusters, type);
System.out.printf("\n\n==================\nFINISHED STAGE 2.\n======================\n\n");
return clusters;
}