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

Bug fix: alignments ending with 'I' were not counted into the overall coverage which resulted in inaccurate stats, and in rare occasions outright messed up ones. 

git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@2635 348d0f76-0448-11de-a6fe-93d51630548a
This commit is contained in:
asivache 2010-01-19 22:12:16 +00:00
parent 96a053c769
commit 4625261d79
1 changed files with 27 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

32
java/src/org/broadinstitute/sting/gatk/walkers/indels/IndelGenotyperV2Walker.java
View File

@ -615,7 +615,7 @@ public class IndelGenotyperV2Walker extends ReadWalker<Integer,Integer> {
public IndelPrecall(WindowContext context, long position, int nqs_width) {
this.pos = position;
this.nqs = nqs_width;
total_coverage = context.coverageAt(pos);
total_coverage = context.coverageAt(pos,true);
List<IndelVariant> variants = context.indelsAt(pos);
findConsensus(variants);
@ -962,13 +962,35 @@ public class IndelGenotyperV2Walker extends ReadWalker<Integer,Integer> {
public Set<ExpandedSAMRecord> getReads() { return reads; }
/** Returns the number of reads spanning over the specified reference position
* (regardless of whether they have a base or indel at that specific location)
* @param refPos position on the reference; must be within the bounds of the window
* (regardless of whether they have a base or indel at that specific location).
* The second argument controls whether to count with indels in mind (this is relevant for insertions only,
* deletions do not require any special treatment since they occupy non-zero length on the ref and since
* alignment can not start or end with a deletion). For insertions, note that, internally, we assign insertions
* to the reference position right after the actual event, and we count all events assigned to a given position.
* This count (reads with indels) should be contrasted to reads without indels, or more rigorously, reads
* that support the ref rather than the indel. Few special cases may occur here:
* 1) an alignment that ends (as per getAlignmentEnd()) right before the current position but has I as its
* last element: we have to count that read into the "coverage" at the current position for the purposes of indel
* assessment, as the indel in that read <i>will</i> be counted at the current position, so the total coverage
* should be consistent with that.
*/
public int coverageAt(final long refPos) {
/* NOT IMPLEMENTED: 2) alsignments that start exactly at the current position do <i>not</i> count for the purpose of insertion
* assessment since they do not contribute any evidence to either Ref or Alt=insertion hypothesis, <i>unless</i>
* the alignment starts with I (so that we do have evidence for an indel assigned to the current position and
* read should be counted). For deletions, reads starting at the current position should always be counted (as they
* show no deletion=ref).
* @param refPos position on the reference; must be within the bounds of the window
*/
public int coverageAt(final long refPos, boolean countForIndels) {
int cov = 0;
for ( ExpandedSAMRecord read : reads ) {
if ( read.getSAMRecord().getAlignmentStart() > refPos || read.getSAMRecord().getAlignmentEnd() < refPos ) continue;
if ( read.getSAMRecord().getAlignmentStart() > refPos || read.getSAMRecord().getAlignmentEnd() < refPos ) {
if ( countForIndels && read.getSAMRecord().getAlignmentEnd() == refPos - 1) {
Cigar c = read.getSAMRecord().getCigar();
if ( c.getCigarElement(c.numCigarElements()-1).getOperator() == CigarOperator.I ) cov++;
}
continue;
}
cov++;
}
return cov;