zzh
/
gatk-3.8
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Hooked up both optimization models via command line arguments. 

git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@2955 348d0f76-0448-11de-a6fe-93d51630548a
This commit is contained in:
rpoplin 2010-03-08 14:49:59 +00:00
parent cc6a714c09
commit fe8a8b9199
3 changed files with 46 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

24
java/src/org/broadinstitute/sting/playground/gatk/walkers/variantoptimizer/VariantNearestNeighborsModel.java
View File

@ -35,8 +35,11 @@ import java.io.PrintStream;
public final class VariantNearestNeighborsModel extends VariantOptimizationModel {
public VariantNearestNeighborsModel( VariantDataManager _dataManager, final double _targetTITV ) {
super( _dataManager, _targetTITV );
private final int numKNN;
public VariantNearestNeighborsModel( VariantDataManager _dataManager, final double _targetTITV, final int _numKNN ) {
super( _dataManager, _targetTITV );
numKNN = _numKNN;
}
public void run( final String outputPrefix ) {
@ -45,7 +48,7 @@ public final class VariantNearestNeighborsModel extends VariantOptimizationModel
final double[] pTrueVariant = new double[numVariants];
final VariantTree vTree = new VariantTree( 2000 );
final VariantTree vTree = new VariantTree( numKNN );
vTree.createTreeFromData( dataManager.data );
System.out.println("Finished creating the kd-tree.");
@ -54,7 +57,18 @@ public final class VariantNearestNeighborsModel extends VariantOptimizationModel
pTrueVariant[iii] = calcTruePositiveRateFromTITV( vTree.calcNeighborhoodTITV( dataManager.data[iii] ) );
}
//BUGBUG: need to output pTrueVariant and other metrics in this method
//return pTrueVariant;
PrintStream outputFile = null;
try {
outputFile = new PrintStream( outputPrefix + ".knn.optimize" );
} catch (Exception e) {
e.printStackTrace();
System.exit(-1);
}
for(int iii = 0; iii < numVariants; iii++) {
outputFile.print(String.format("%.4f",pTrueVariant[iii]) + ",");
outputFile.println( (dataManager.data[iii].isTransition ? 1 : 0)
+ "," + (dataManager.data[iii].isKnown? 1 : 0)
+ "," + (dataManager.data[iii].isFiltered ? 1 : 0) );
}
}
}

6
java/src/org/broadinstitute/sting/playground/gatk/walkers/variantoptimizer/VariantOptimizationModel.java
View File

@ -32,6 +32,12 @@ package org.broadinstitute.sting.playground.gatk.walkers.variantoptimizer;
*/
public abstract class VariantOptimizationModel implements VariantOptimizationInterface {
public enum Model {
GAUSSIAN_MIXTURE_MODEL,
K_NEAREST_NEIGHBORS
}
protected final VariantDataManager dataManager;
protected final double targetTITV;

26
java/src/org/broadinstitute/sting/playground/gatk/walkers/variantoptimizer/VariantOptimizer.java
View File

@ -7,6 +7,7 @@ import org.broadinstitute.sting.gatk.refdata.RefMetaDataTracker;
import org.broadinstitute.sting.gatk.walkers.RodWalker;
import org.broadinstitute.sting.utils.BaseUtils;
import org.broadinstitute.sting.utils.ExpandingArrayList;
import org.broadinstitute.sting.utils.StingException;
import org.broadinstitute.sting.utils.cmdLine.Argument;
/*
@ -61,10 +62,14 @@ public class VariantOptimizer extends RodWalker<ExpandingArrayList<VariantDatum>
private String[] FORCED_ANNOTATIONS = null;
@Argument(fullName="output", shortName="output", doc="The output file name", required=false)
private String OUTPUT_FILE = "optimizer.data";
@Argument(fullName="numGaussians", shortName="nG", doc="The number of Gaussians to be used in the Gaussian mixture model", required=false)
@Argument(fullName="numGaussians", shortName="nG", doc="The number of Gaussians to be used in the Gaussian Mixture model", required=false)
private int NUM_GAUSSIANS = 32;
@Argument(fullName="numIterations", shortName="nI", doc="The number of iterations to be performed in the Gaussian mixture model", required=false)
private int NUM_ITERATIONS = 5; //BUGBUG: should automatically decided when to stop by looking at how entropy changes with each iteration
@Argument(fullName="numIterations", shortName="nI", doc="The number of iterations to be performed in the Gaussian Mixture model", required=false)
private int NUM_ITERATIONS = 10;
@Argument(fullName="knn", shortName="knn", doc="The number of nearest neighbors to be used in the k-Nearest Neighbors model", required=false)
private int NUM_KNN = 2000;
@Argument(fullName = "optimization_model", shortName = "om", doc = "Optimization calculation model to employ -- GAUSSIAN_MIXTURE_MODEL is currently the default, while K_NEAREST_NEIGHBORS is also available for small callsets.", required = false)
private VariantOptimizationModel.Model OPTIMIZATION_MODEL = VariantOptimizationModel.Model.GAUSSIAN_MIXTURE_MODEL;
/////////////////////////////
@ -179,8 +184,19 @@ public class VariantOptimizer extends RodWalker<ExpandingArrayList<VariantDatum>
dataManager.normalizeData(); // Each data point is now [ (x - mean) / standard deviation ]
// Create either the Gaussian Mixture Model or the Nearest Neighbors model and run it
final VariantOptimizationModel gmm = new VariantGaussianMixtureModel( dataManager, TARGET_TITV, NUM_GAUSSIANS, NUM_ITERATIONS );
gmm.run( OUTPUT_FILE );
VariantOptimizationModel theModel;
switch (OPTIMIZATION_MODEL) {
case GAUSSIAN_MIXTURE_MODEL:
theModel = new VariantGaussianMixtureModel( dataManager, TARGET_TITV, NUM_GAUSSIANS, NUM_ITERATIONS );
break;
case K_NEAREST_NEIGHBORS:
theModel = new VariantNearestNeighborsModel( dataManager, TARGET_TITV, NUM_KNN );
break;
default:
throw new StingException("Variant Optimization Model is unrecognized. Implemented options are GAUSSIAN_MIXTURE_MODEL and K_NEAREST_NEIGHBORS");
}
theModel.run( OUTPUT_FILE );
}
}