[Samtools-help] CollectAlignmentSummaryMetrics reports zero aligned reads

[Jeremy V. <vol...@wi...>]

Hello,

I have a set of ~ 1 billion gDNA paired reads mapped to a reference
genome with BWA-MEM. I tried to use picard's
CollectAlignmentSummaryMetrics to generate a summary of the mapping,
but it reports zero aligned reads:

FIRST_OF_PAIR   549154663   549154663   1   0   0   0   0   0   0   0   0   0   0   0   99.936337   0   0   0   0   0   0.000001
SECOND_OF_PAIR  549154663   549154663   1   0   0   0   0   0   0   0   0   0   0   0   98.611736   0   0   0   0   0   0
PAIR    1098309326  1098309326  1   0   0   0   0   0   0   0   0   0   0   0   99.274036   0   0   0   0   0   0

A look at the SAM flags manually seems to indicate that most pairs are properly aligned,
and the output of samtools flagstat seems to agree:

1102426754 + 0 in total (QC-passed reads + QC-failed reads)
0 + 0 duplicates
1097354292 + 0 mapped (99.54%:-nan%)
1102426754 + 0 paired in sequencing
551276790 + 0 read1
551149964 + 0 read2
1042528490 + 0 properly paired (94.57%:-nan%)
1096261806 + 0 with itself and mate mapped
1092486 + 0 singletons (0.10%:-nan%)
36064341 + 0 with mate mapped to a different chr 
14493831 + 0 with mate mapped to a different chr (mapQ>=5)

I don't mind using samtools flagstat to evaluate the mapping, but I'm puzzled by this
and concerned that the problems with the picard output signal further potential issues
using picard for duplicate removal, etc. I've seen a few users report similar behavior
in the past, and one was able to pin it to their combination of OS and java, but I've
tried multiple combinations of java package (Sun, OpenJDK) and picard version and get
the same results.

To generate the above results I used:

Debian 'Wheezy'
java version 1.7.0_67
picard version 1.119
samtools version 0.1.18

and the picard command was:

java -jar /opt/picard/CollectAlignmentSummaryMetrics.jar MAX_INSERT_SIZE=500 I=bwa.map.w_rg.bam O=bwa.picard.summary

As per BWA the library insert size distribution has mean and sd of ~ 180 and 20, respectively.
I am attaching a pruned SAM file with just a pair of reads that appear to be mapped correctly
but which picard reports as unaligned as per above. Any suggestions or help would be much
appreciated.

Thanks,
Jeremy