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

Swapping the iteration order to run over AF conformations and then samples instead of the reverse minimizes calls to HashMap.get; instead of it being O(n) since we called it for each sample it's now O(1). Runtime on T2D GENES test set is reduced by 5-10%. More optimizations to follow.

This commit is contained in:
Eric Banks 2012-01-12 02:04:03 -05:00
parent 34cf2fe43b
commit 410a340ef5
1 changed files with 39 additions and 32 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

71
public/java/src/org/broadinstitute/sting/gatk/walkers/genotyper/ExactAFCalculationModel.java
View File

@ -35,7 +35,7 @@ import java.util.*;
public class ExactAFCalculationModel extends AlleleFrequencyCalculationModel {
private final static boolean DEBUG = false;
// private final static boolean DEBUG = false;
private final static double MAX_LOG10_ERROR_TO_STOP_EARLY = 6; // we want the calculation to be accurate to 1 / 10^6
@ -199,8 +199,8 @@ public class ExactAFCalculationModel extends AlleleFrequencyCalculationModel {
final double[][] log10AlleleFrequencyPriors,
final AlleleFrequencyCalculationResult result) {
if ( DEBUG )
System.out.printf(" *** computing LofK for set=%s%n", set.ACcounts);
//if ( DEBUG )
// System.out.printf(" *** computing LofK for set=%s%n", set.ACcounts);
// compute the log10Likelihoods
computeLofK(set, genotypeLikelihoods, indexesToACset, log10AlleleFrequencyPriors, result);
@ -209,8 +209,8 @@ public class ExactAFCalculationModel extends AlleleFrequencyCalculationModel {
if ( !preserveData ) {
for ( ExactACcounts index : set.dependentACsetsToDelete ) {
indexesToACset.put(index, null);
if ( DEBUG )
System.out.printf(" *** removing used set=%s after seeing final dependent set=%s%n", index, set.ACcounts);
//if ( DEBUG )
// System.out.printf(" *** removing used set=%s after seeing final dependent set=%s%n", index, set.ACcounts);
}
}
@ -218,8 +218,8 @@ public class ExactAFCalculationModel extends AlleleFrequencyCalculationModel {
// can we abort early because the log10Likelihoods are so small?
if ( log10LofK < maxLog10L - MAX_LOG10_ERROR_TO_STOP_EARLY ) {
if ( DEBUG )
System.out.printf(" *** breaking early set=%s log10L=%.2f maxLog10L=%.2f%n", set.ACcounts, log10LofK, maxLog10L);
//if ( DEBUG )
// System.out.printf(" *** breaking early set=%s log10L=%.2f maxLog10L=%.2f%n", set.ACcounts, log10LofK, maxLog10L);
// no reason to keep this data around because nothing depends on it
if ( !preserveData )
@ -262,8 +262,8 @@ public class ExactAFCalculationModel extends AlleleFrequencyCalculationModel {
// if the last dependent set was not at the back of the queue (i.e. not just added), then we need to iterate
// over all the dependent sets to find the last one in the queue (otherwise it will be cleaned up too early)
if ( !preserveData && lastSet == null ) {
if ( DEBUG )
System.out.printf(" *** iterating over dependent sets for set=%s%n", set.ACcounts);
//if ( DEBUG )
// System.out.printf(" *** iterating over dependent sets for set=%s%n", set.ACcounts);
lastSet = determineLastDependentSetInQueue(set.ACcounts, ACqueue);
}
if ( lastSet != null )
@ -323,36 +323,43 @@ public class ExactAFCalculationModel extends AlleleFrequencyCalculationModel {
else {
// all possible likelihoods for a given cell from which to choose the max
final int numPaths = set.ACsetIndexToPLIndex.size() + 1;
final double[] log10ConformationLikelihoods = new double[numPaths]; // TODO can be created just once, since you initialize it
final double[][] log10ConformationLikelihoods = new double[set.log10Likelihoods.length][numPaths]; // TODO can be created just once, since you initialize it
// initialize
for ( int i = 0; i < set.log10Likelihoods.length; i++ )
for ( int j = 0; j < numPaths; j++ )
// TODO -- Arrays.fill?
// todo -- is this even necessary? Why not have as else below?
log10ConformationLikelihoods[i][j] = Double.NEGATIVE_INFINITY;
for ( int j = 1; j < set.log10Likelihoods.length; j++ ) {
final double[] gl = genotypeLikelihoods.get(j);
final double logDenominator = MathUtils.log10Cache[2*j] + MathUtils.log10Cache[2*j-1];
// deal with the non-AA possible conformations
int conformationIndex = 1;
for ( Map.Entry<ExactACcounts, Integer> mapping : set.ACsetIndexToPLIndex.entrySet() ) {
//if ( DEBUG )
// System.out.printf(" *** evaluating set=%s which depends on set=%s%n", set.ACcounts, mapping.getKey());
// initialize
for ( int i = 0; i < numPaths; i++ )
// TODO -- Arrays.fill?
// todo -- is this even necessary? Why not have as else below?
log10ConformationLikelihoods[i] = Double.NEGATIVE_INFINITY;
ExactACset dependent = indexesToACset.get(mapping.getKey());
// deal with the AA case first
if ( totalK < 2*j-1 )
log10ConformationLikelihoods[0] = MathUtils.log10Cache[2*j-totalK] + MathUtils.log10Cache[2*j-totalK-1] + set.log10Likelihoods[j-1] + gl[HOM_REF_INDEX];
for ( int j = 1; j < set.log10Likelihoods.length; j++ ) {
final double[] gl = genotypeLikelihoods.get(j);
// deal with the other possible conformations now
if ( totalK <= 2*j ) { // skip impossible conformations
int conformationIndex = 1;
for ( Map.Entry<ExactACcounts, Integer> mapping : set.ACsetIndexToPLIndex.entrySet() ) {
if ( DEBUG )
System.out.printf(" *** evaluating set=%s which depends on set=%s%n", set.ACcounts, mapping.getKey());
log10ConformationLikelihoods[conformationIndex++] =
determineCoefficient(mapping.getValue(), j, set.ACcounts.getCounts(), totalK) + indexesToACset.get(mapping.getKey()).log10Likelihoods[j-1] + gl[mapping.getValue()];
if ( totalK <= 2*j ) { // skip impossible conformations
log10ConformationLikelihoods[j][conformationIndex] =
determineCoefficient(mapping.getValue(), j, set.ACcounts.getCounts(), totalK) + dependent.log10Likelihoods[j-1] + gl[mapping.getValue()];
}
}
final double log10Max = MathUtils.approximateLog10SumLog10(log10ConformationLikelihoods);
conformationIndex++;
}
// finally, update the L(j,k) value
// finally, deal with the AA case (which depends on previous cells in this column) and then update the L(j,k) value
for ( int j = 1; j < set.log10Likelihoods.length; j++ ) {
final double[] gl = genotypeLikelihoods.get(j);
if ( totalK < 2*j-1 )
log10ConformationLikelihoods[j][0] = MathUtils.log10Cache[2*j-totalK] + MathUtils.log10Cache[2*j-totalK-1] + set.log10Likelihoods[j-1] + gl[HOM_REF_INDEX];
final double logDenominator = MathUtils.log10Cache[2*j] + MathUtils.log10Cache[2*j-1];
final double log10Max = MathUtils.approximateLog10SumLog10(log10ConformationLikelihoods[j]);
set.log10Likelihoods[j] = log10Max - logDenominator;
}
}
@ -523,7 +530,7 @@ public class ExactAFCalculationModel extends AlleleFrequencyCalculationModel {
lastK = k;
maxLog10L = Math.max(maxLog10L, log10LofK);
if ( log10LofK < maxLog10L - MAX_LOG10_ERROR_TO_STOP_EARLY ) {
if ( DEBUG ) System.out.printf(" *** breaking early k=%d log10L=%.2f maxLog10L=%.2f%n", k, log10LofK, maxLog10L);
//if ( DEBUG ) System.out.printf(" *** breaking early k=%d log10L=%.2f maxLog10L=%.2f%n", k, log10LofK, maxLog10L);
done = true;
}