Problem with QC metrics

[froggins]

I have realigned, recalibrated BAM files on which I am trying to run final QC metrics. I have tried running HS metrics and AlignmentSumaryMetrics and I get the following error messages.

1. CalculateHSMetrics - "Sequence dictionaries are not the same size (0, 25)"
Does this refer to the difference between my reference file and BMA file? Do I need to create a BAM Sequence dictionary for my reference file (in fasta format)?

[Wed Jan 05 11:06:13 GMT 2011] net.sf.picard.analysis.directed.CalculateHsMetrics done.
Runtime.totalMemory()=2058027008
Exception in thread "main" net.sf.samtools.util.SequenceUtil$SequenceListsDif ferException: Sequence dictionaries are not the same size (0, 25)
at net.sf.samtools.util.SequenceUtil.assertSequenceLi stsEqual(SequenceUtil.java:76)
at net.sf.samtools.util.SequenceUtil.assertSequenceDi ctionariesEqual(SequenceUtil.java:118)
at net.sf.picard.analysis.directed.CalculateHsMetrics .doWork(CalculateHsMetrics.java:87)
at net.sf.picard.cmdline.CommandLineProgram.instanceM ain(CommandLineProgram.java:156)
at net.sf.picard.analysis.directed.CalculateHsMetrics .main(CalculateHsMetrics.java:68)


2. Alignment Summary Metrics - myfile.bam has no sequence dictionary
Again, does this mean that I need to make a separate sequence dictionary associated with myfile.bam? Presumably this is the source of the error above. Is it possible to generate a sequence dictionary from a bam file with Picard? It also says "If any reads in the file are aligned then alignment summary metrics collection will fail" (which I found confusing because I thought that the file was supposed to be aligned).


WARNING 2011-01-05 11:36:59 CollectAlignmentSummaryMetrics myfile.bam has no sequence dictionary. If any reads in the file are aligned then alignment summary metrics collection will fail.
[Wed Jan 05 11:37:00 GMT 2011] net.sf.picard.analysis.CollectAlignmentSummaryMetr ics done.
Runtime.totalMemory()=2058027008


I am sorry if these are really simple questions - I am totally new to all of these programs. I have tried searching through previous help questions on samtools and here . Part of the problem is that I can't work out exactly what a "sequence dictionary" is (other than a header file - which I am sure is so obvious that it doesn't require explaining!)

If anyone can help I would be very appreciative!

Thanks.

[Seq84]

Hi all,

I have the same problem at point 1.

Thanks in advance for any suggestions.

[froggins]

Hi Seq84,
I ended up getting help from Alec Wysoker on the Samtools help mailing list.

The problem was that my BAM files were incorrectly formed (and therefore did not have sequence dictionaries). In my case, the reason why they were faulty was because my version of GATK used -O as the output command whereas the newer version and description on the GATK website used -o. I had to upgrade my version of GATK and re-align my BAM files first . It then worked fine.

You can check your BAM headers using samtools to make sure that they are correctly formatted.

This is from Alec's reply to me:
CalculateHsMetrics checks that the sequence dictionary in your BAM file
and in your TARGET_INTERVALS file agree, and it appears that they do
not. The message should be more informative about where these come
from. Anyway, the problem is that your BAM has an empty sequence
dictionary, i.e. its size is 0. Are you sure that your BAM is aligned?

Likewise, CollectAlignmentSummaryMetrics doesn't see a sequence
dictionary in your BAM. An aligned BAM must have a non-empty sequence
dictionary. For some reason, yours doesn't. You can look at the header
of your BAM using samtools view -H, or Picard ViewSam (just looking at
the first few lines) to see if there are @SQ records in your BAM. If
not, I would suggest looking at the various stages that your data has
gone through, e.g. confirm that the data as produced by the aligner has
a sequence dictionary, then confirm that after recalibration it has a
header, etc.


If that doesn't help I would suggest trying the samtools mailing list/forum to get more help.

Best wishes,

Kirsten