Improvements to the VariantRecalibrator R plots

-- VariantRecalibrator now emits plots with denormlized values (original values) instead of their normalized (x - mu / sigma) which helps to understand the distribution of values that are good and bad

protected/java/src/org/broadinstitute/sting/gatk/walkers/variantrecalibration/VariantDataManager.java

@@ -127,6 +127,22 @@ public class VariantDataManager {
         }
     }
 
+    /**
+     * Convert a normalized point to it's original annotation value
+     *
+     * norm = (orig - mu) / sigma
+     * orig = norm * sigma + mu
+     *
+     * @param normalizedValue the normalized value of the ith annotation
+     * @param annI the index of the annotation value
+     * @return the denormalized value for the annotation
+     */
+    public double denormalizeDatum(final double normalizedValue, final int annI) {
+        final double mu = meanVector[annI];
+        final double sigma = varianceVector[annI];
+        return normalizedValue * sigma + mu;
+    }
+
     public void addTrainingSet( final TrainingSet trainingSet ) {
         trainingSets.add( trainingSet );
     }

protected/java/src/org/broadinstitute/sting/gatk/walkers/variantrecalibration/VariantRecalibrator.java

@@ -435,14 +435,20 @@ public class VariantRecalibrator extends RodWalker<ExpandingArrayList<VariantDat
 
                 stream.print("surface <- c(");
                 for( final VariantDatum datum : fakeData ) {
-                    stream.print(String.format("%.3f, %.3f, %.3f, ", datum.annotations[iii], datum.annotations[jjj], Math.min(4.0, Math.max(-4.0, datum.lod))));
+                    stream.print(String.format("%.3f, %.3f, %.3f, ",
+                            dataManager.denormalizeDatum(datum.annotations[iii], iii),
+                            dataManager.denormalizeDatum(datum.annotations[jjj], jjj),
+                            Math.min(4.0, Math.max(-4.0, datum.lod))));
                 }
                 stream.println("NA,NA,NA)");
                 stream.println("s <- matrix(surface,ncol=3,byrow=T)");
 
                 stream.print("data <- c(");
                 for( final VariantDatum datum : randomData ) {
-                    stream.print(String.format("%.3f, %.3f, %.3f, %d, %d,", datum.annotations[iii], datum.annotations[jjj], (datum.lod < lodCutoff ? -1.0 : 1.0),
+                    stream.print(String.format("%.3f, %.3f, %.3f, %d, %d,",
+                            dataManager.denormalizeDatum(datum.annotations[iii], iii),
+                            dataManager.denormalizeDatum(datum.annotations[jjj], jjj),
+                            (datum.lod < lodCutoff ? -1.0 : 1.0),
                             (datum.atAntiTrainingSite ? -1 : (datum.atTrainingSite ? 1 : 0)), (datum.isKnown ? 1 : -1)));
                 }
                 stream.println("NA,NA,NA,NA,1)");