Wrong filtering in Alignment metrics #943

Ksenomorfa · 2017-09-29T11:35:19Z

Description

Comment in AlignmentSummaryMetricsCollector method collectReadData() says: " // NB: for read count metrics, do not include supplementary records, but for base count metrics, do include supplementary records". But supplementary records are pre-filtered at super level in method acceptRecord() in the same class, because only non-secondary and non-supplementary records are accepted issue.

Implementation

According to the issue:

Added check for supplementary flag before increment histogram for metrics.BAD_CYCLES in method collectQualityData();
Changed docs of BAD_CYCLES metric in AlignmentSummaryMetrics;
Created testSupplementaryReadsAndBases() and tested SAM-file containing supplementary alignment.

Checklist (never delete this)

Never delete this, it is our record that procedure was followed. If you find that for whatever reason one of the checklist points doesn't apply to your PR, you can leave it unchecked but please add an explanation below.

Content

Added or modified tests to cover changes and any new functionality
Edited the README / documentation (if applicable)
All tests passing on Travis

Review

Final thumbs-up from reviewer
Rebase, squash and reword as applicable

For more detailed guidelines, see https://github.com/broadinstitute/picard/wiki/Guidelines-for-pull-requests

coveralls · 2017-09-29T11:55:36Z

Coverage increased (+0.006%) to 76.775% when pulling 2b99ffb on EpamLifeSciencesTeam:epam-ls_WrongFilteringInAlignmentMetrics into d4b04a9 on broadinstitute:master.

yfarjoun · 2017-09-29T12:19:39Z

@nh13, I think that you are responsible for these NB's....will have time to take a look at these changes?

nh13

Great catch! Some discussion about getSupplementaryAlignmentFlag is still needed.

nh13 · 2017-10-16T17:33:11Z

src/main/java/picard/analysis/AlignmentSummaryMetricsCollector.java

@@ -270,7 +268,8 @@ private void collectQualityData(final SAMRecord record, final ReferenceSequence
                // NB: for read count metrics, do not include supplementary records, but for base count metrics, do include supplementary records.

                // If the read isn't an aligned PF read then look at the read for no-calls
-                if (record.getReadUnmappedFlag() || record.getReadFailsVendorQualityCheckFlag() || !doRefMetrics) {
+                if (!record.getSupplementaryAlignmentFlag() &&


Why is this check added getSupplementaryAlignmentFlag? I thought we wanted to have supplementary alignments add to these metrics (ex. PF_HQ_MEDIAN_MISMATCHES)? @yfarjoun want to weigh in here?

I think that the idea is that a supplementary read should not be considered for the first part (namely updating badCycles histogram on unmapped reads) but, it should be considered for the "else" part....I agree with that sentiment, but I also agree with you that it wasn't entirely well thought out, in particular: a supplementary alignment will cause the code to enter the second part even all the other conditions (mapped, passesFilter, refMetrics) are false.

Done, changed if-else condition, sorry for this omission.

nh13 · 2017-10-16T17:33:16Z

src/main/java/picard/analysis/AlignmentSummaryMetricsCollector.java

@@ -314,7 +313,8 @@ else if (!record.getReadFailsVendorQualityCheckFlag()) {
                                if (mismatch) hqMismatchCount++;
                            }

-                            if (mismatch || SequenceUtil.isNoCall(readBases[readBaseIndex])) {
+                            if (!record.getSupplementaryAlignmentFlag()


If I understand correctly, we can leave this "if" condition for supplementary reads (according to yfarjoun comment)?

I think that this is right.. @nh13 what do you think?

ah no. sorry. this is more complex. Given that we are only looping over the alignment blocks, we need to look at the supplemental reads in order to see the bases....I guess, it's as @nh13 said in the code comment: for metrics that have to do with bases (so the badCycleHistogram is one), we want to see the supplemental reads, but for metrics that have to do with reads, we do not. I think that this means that supplemental reads should be filtered from collectReadData (if supplemental, return) but not from collectQualityData.

Thank you for clarifying this! Okay, so I want to make sure that I understand right.. :) I can remove all filtering for supplementary from collectQualityData() , and leave check for secondary in acceptRecord() - this will be necessary logic?

I think that's right. @nh13 ?

yfarjoun · 2017-10-16T19:26:55Z

src/main/java/picard/analysis/AlignmentSummaryMetricsCollector.java

@@ -79,9 +79,7 @@ public AlignmentSummaryMetricsCollector(final Set<MetricAccumulationLevel> accum

    @Override
    public void acceptRecord(final SAMRecord rec, final ReferenceSequence ref) {
-        if (!rec.isSecondaryOrSupplementary()) {


Do we want to include secondary reads as well? I think not. we should probably filter secondary reads here. Even if they are filtered upstream somehow (I don't think that they are) this program should be clear that secondary alignements are filtered out.

Done, added check for secondary read here.

yfarjoun · 2017-10-16T20:23:12Z

src/main/java/picard/analysis/AlignmentSummaryMetricsCollector.java

@@ -270,7 +268,8 @@ private void collectQualityData(final SAMRecord record, final ReferenceSequence
                // NB: for read count metrics, do not include supplementary records, but for base count metrics, do include supplementary records.

                // If the read isn't an aligned PF read then look at the read for no-calls
-                if (record.getReadUnmappedFlag() || record.getReadFailsVendorQualityCheckFlag() || !doRefMetrics) {
+                if (!record.getSupplementaryAlignmentFlag() &&


I think that the idea is that a supplementary read should not be considered for the first part (namely updating badCycles histogram on unmapped reads) but, it should be considered for the "else" part....I agree with that sentiment, but I also agree with you that it wasn't entirely well thought out, in particular: a supplementary alignment will cause the code to enter the second part even all the other conditions (mapped, passesFilter, refMetrics) are false.

* Fix if-else condition of supplementary reads

yfarjoun · 2017-10-23T09:03:03Z

testdata/picard/sam/summary_alignment_stats_test_supplementary.sam

+@SQ	SN:chr6	LN:101
+@SQ	SN:chr8	LN:202
+@RG	ID:0	SM:Hi,Momma!	LB:whatever	PU:me	PL:ILLUMINA
+SL-XAV:1:1:0:700#0/2	137	chr6	1	255	101M	*	0	0	NAATTGTTCTNAGTTTCTCGGTTTATGTGCTCTTCCAGGTGGGTAACACAATAATGGCCTTCCAGATCGTAAGAGCGACGTGTGTTGCACNAGTGTCGATC	&0::887::::6/646::838388811/679:87640&./2+/-4/28:3,536/4''&&.78/(/554/./02*)*',-(57()&.6(6:(0601'/(,*	RG:Z:0


it would be good if this example were more realistic, so that the primary read were only aligned for, say 60 bases and the supplementary read would be aligned for the remaining 41 bases. also, the bases and the qualities should be the same for the read and its supplemental alignment....

@PolinaBevad are you going to add a more realistic test?

Wrong filtering in Alignment metrics

2b99ffb

yfarjoun requested a review from nh13 September 29, 2017 11:39

nh13 requested changes Oct 16, 2017

View reviewed changes

yfarjoun suggested changes Oct 22, 2017

View reviewed changes

* Add check for secondary reads

1f4fbc8

* Fix if-else condition of supplementary reads

yfarjoun reviewed Oct 23, 2017

View reviewed changes

yfarjoun assigned gbggrant Jun 6, 2018

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Wrong filtering in Alignment metrics #943

Wrong filtering in Alignment metrics #943

Ksenomorfa commented Sep 29, 2017

coveralls commented Sep 29, 2017 •

edited

Loading

yfarjoun commented Sep 29, 2017

nh13 left a comment

nh13 Oct 16, 2017

yfarjoun Oct 16, 2017

Ksenomorfa Oct 23, 2017

nh13 Oct 16, 2017

Ksenomorfa Oct 23, 2017 •

edited

Loading

yfarjoun Oct 23, 2017

yfarjoun Oct 23, 2017

Ksenomorfa Oct 23, 2017

yfarjoun Oct 23, 2017

yfarjoun Oct 16, 2017

Ksenomorfa Oct 23, 2017

yfarjoun Oct 16, 2017

yfarjoun Oct 23, 2017

yfarjoun Dec 11, 2018

Wrong filtering in Alignment metrics #943

Are you sure you want to change the base?

Wrong filtering in Alignment metrics #943

Conversation

Ksenomorfa commented Sep 29, 2017

Description

Implementation

Checklist (never delete this)

Content

Review

coveralls commented Sep 29, 2017 • edited Loading

yfarjoun commented Sep 29, 2017

nh13 left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Ksenomorfa Oct 23, 2017 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

coveralls commented Sep 29, 2017 •

edited

Loading

Ksenomorfa Oct 23, 2017 •

edited

Loading