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

Added histogram of variant quality scores broken out by true positive and false positive calls to the GenotypeConcordance module of VariantEval2 

git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@3123 348d0f76-0448-11de-a6fe-93d51630548a
This commit is contained in:
rpoplin 2010-04-06 13:48:31 +00:00
parent 12e4f88ca7
commit 2d002c56c3
1 changed files with 142 additions and 20 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

162
java/src/org/broadinstitute/sting/oneoffprojects/walkers/varianteval2/GenotypeConcordance.java
View File

@ -23,7 +23,7 @@ import java.util.*;
*/
@Analysis(name = "Genotype Concordance", description = "Determine the genotype concordance between the genotypes in difference tracks")
public class GenotypeConcordance extends VariantEvaluator {
protected static Logger logger = Logger.getLogger(GenotypeConcordance.class);
protected final static Logger logger = Logger.getLogger(GenotypeConcordance.class);
// a mapping from allele count to stats
@DataPoint(description = "the frequency statistics for each allele")
@ -33,6 +33,9 @@ public class GenotypeConcordance extends VariantEvaluator {
@DataPoint(name="samples", description = "the concordance statistics for each sample")
SampleStats sampleStats = null;
@DataPoint(description = "the variant quality score histograms for true positive and false positive calls")
QualityScoreHistograms qualityScoreHistograms = null;
private static final int MAX_MISSED_VALIDATION_DATA = 10000;
@ -63,6 +66,93 @@ public class GenotypeConcordance extends VariantEvaluator {
}
}
class QualityScoreHistograms implements TableType {
final int NUM_BINS = 20;
final ArrayList<Double> truePositiveQualities = new ArrayList<Double>(); // An ArrayList holds all the qualities until we are able to bin them appropriately
final ArrayList<Double> falsePositiveQualities = new ArrayList<Double>();
final int truePositiveHist[] = new int[NUM_BINS]; // the final histograms that get reported out
final int falsePositiveHist[] = new int[NUM_BINS];
final String[] rowKeys = new String[]{"true_positive_hist", "false_positive_hist"};
public Object[] getRowKeys() {
return rowKeys;
}
public Object[] getColumnKeys() {
final String columnKeys[] = new String[NUM_BINS];
for( int iii = 0; iii < NUM_BINS; iii++ ) {
columnKeys[iii] = "histBin" + iii;
}
return columnKeys;
}
public String getName() {
return "QualityScoreHistogram";
}
public String getCell(int x, int y) {
if( x == 0 ) {
return String.valueOf(truePositiveHist[y]);
} else if ( x == 1 ) {
return String.valueOf(falsePositiveHist[y]);
} else {
throw new StingException( "Unknown row in " + getName() + ", row = " + x );
}
}
public String toString() {
String returnString = "";
// output both histogram arrays
returnString += "TP: ";
for( int iii = 0; iii < NUM_BINS; iii++ ) {
returnString += truePositiveHist[iii] + " ";
}
returnString += "\nFP: ";
for( int iii = 0; iii < NUM_BINS; iii++ ) {
returnString += falsePositiveHist[iii] + " ";
}
return returnString;
}
public void incrValue( final double qual, final boolean isTruePositiveCall ) {
if( isTruePositiveCall ) {
truePositiveQualities.add( qual );
} else {
falsePositiveQualities.add( qual );
}
}
public void organizeHistogramTables() {
for( int iii = 0; iii < NUM_BINS; iii++ ) {
truePositiveHist[iii] = 0;
falsePositiveHist[iii] = 0;
}
double maxQual = 0.0;
// Calculate the maximum quality score for both TP and FP calls in order to normalize and histogram
for( final Double qual : truePositiveQualities ) {
if( qual > maxQual ) {
maxQual = qual;
}
}
for( final Double qual : falsePositiveQualities ) {
if( qual > maxQual ) {
maxQual = qual;
}
}
final double binSize = maxQual / ((double) (NUM_BINS-1));
for( final Double qual : truePositiveQualities ) {
truePositiveHist[ (int)Math.floor( qual / binSize ) ]++;
}
for( final Double qual : falsePositiveQualities ) {
falsePositiveHist[ (int)Math.floor( qual / binSize ) ]++;
}
}
}
// keep a list of the validation data we saw before the first eval data
private HashSet<VariantContext> missedValidationData = new HashSet<VariantContext>();
@ -115,18 +205,24 @@ public class GenotypeConcordance extends VariantEvaluator {
private boolean warnedAboutValidationData = false;
public String update2(VariantContext eval, VariantContext validation, RefMetaDataTracker tracker, ReferenceContext ref, AlignmentContext context) {
String interesting = null;
final String interesting = null;
// sanity check that we at least have either eval or validation data
if (eval == null && !isValidVC(validation))
if (eval == null && !isValidVC(validation)) {
return interesting;
}
if( qualityScoreHistograms == null ) {
qualityScoreHistograms = new QualityScoreHistograms();
}
if (sampleStats == null) {
if (eval != null) {
// initialize the concordance table
sampleStats = new SampleStats(eval,Genotype.Type.values().length);
for (VariantContext vc : missedValidationData)
for (final VariantContext vc : missedValidationData) {
determineStats(null, vc);
}
missedValidationData = null;
} else {
// todo -- Eric, this results in a memory problem when eval is WEx data but you are using CG calls genome-wide
@ -149,51 +245,77 @@ public class GenotypeConcordance extends VariantEvaluator {
return interesting; // we don't capture any interesting sites
}
private void determineStats(VariantContext eval, VariantContext validation) {
private void determineStats(final VariantContext eval, final VariantContext validation) {
final boolean validationIsValidVC = isValidVC(validation);
// determine concordance for eval data
if (eval != null) {
for (String sample : eval.getSampleNames()) {
Genotype.Type called = eval.getGenotype(sample).getType();
Genotype.Type truth;
for (final String sample : eval.getSampleNames()) {
final Genotype.Type called = eval.getGenotype(sample).getType();
final Genotype.Type truth;
if (!isValidVC(validation) || !validation.hasGenotype(sample))
if (!validationIsValidVC || !validation.hasGenotype(sample)) {
truth = Genotype.Type.NO_CALL;
else
} else {
truth = validation.getGenotype(sample).getType();
}
sampleStats.incrValue(sample, truth, called);
}
}
// otherwise, mark no-calls for all samples
else {
Genotype.Type called = Genotype.Type.NO_CALL;
final Genotype.Type called = Genotype.Type.NO_CALL;
for (String sample : validation.getSampleNames()) {
Genotype.Type truth = validation.getGenotype(sample).getType();
for (final String sample : validation.getSampleNames()) {
final Genotype.Type truth = validation.getGenotype(sample).getType();
sampleStats.incrValue(sample, truth, called);
}
}
// determine allele count concordance ()
if (isValidVC(validation) && validation.isPolymorphic()) {
if (validationIsValidVC && validation.isPolymorphic()) {
int trueAlleleCount = 0;
for (Allele a : validation.getAlternateAlleles())
for (final Allele a : validation.getAlternateAlleles()) {
trueAlleleCount += validation.getChromosomeCount(a);
}
if (!alleleCountStats.containsKey(trueAlleleCount))
if (!alleleCountStats.containsKey(trueAlleleCount)) {
alleleCountStats.put(trueAlleleCount, new FrequencyStats());
FrequencyStats stats = alleleCountStats.get(trueAlleleCount);
if (eval != null)
}
final FrequencyStats stats = alleleCountStats.get(trueAlleleCount);
if (eval != null) {
stats.nFound++;
else
} else {
stats.nMissed++;
}
}
// TP & FP quality score histograms
if( eval != null && eval.isPolymorphic() && validationIsValidVC ) {
if( eval.getSampleNames().size() == 1 ) { // single sample calls
for( final String sample : eval.getSampleNames() ) { // only one sample
if( validation.hasGenotype(sample) ) {
final Genotype truth = validation.getGenotype(sample);
qualityScoreHistograms.incrValue( eval.getPhredScaledQual(), !truth.isHomRef() );
}
}
} else { // multi sample calls
qualityScoreHistograms.incrValue( eval.getPhredScaledQual(), validation.isPolymorphic() );
}
}
}
private static boolean isValidVC(VariantContext vc) {
private static boolean isValidVC(final VariantContext vc) {
return (vc != null && !vc.isFiltered());
}
public void finalizeEvaluation() {
qualityScoreHistograms.organizeHistogramTables();
}
}
/**