Error in flags after picard FixMateInformation?

[oriol]

Hi!

I am not sure, but it seems that after using Picard's FixMateInformation, some flags of the BAM file are wrong.

For instance, my initial BAM file has these two reads:

HWI-ST731_145:7:1306:17910:58342#15@0 409 12 57997071 3 76M * 0 0 GGAGGAGATCAGCGAGAGGCTGGAAGAAGCTGGCGGGGCGACGTCAGCCCAGATTGAGATGAACAAGAAGCGCGAG DDDDDEDDDDDDDDCDDDEEEECEEDFFFGHHJJJJJJJJJIIIJIJIJJJJIIJJIJJIIJIHHHHHFFFFFCCC CC:Z:= MD:Z:30C45 RG:Z1126ACXX_7_15 XG:i:0 NH:i:2 HI:i:0 NM:i:1 XM:i:1 XN:i:0 XO:i:0 CP:i:58221643 AS:i:-5 XS:A:+ YT:Z:UU
HWI-ST731_145:7:1306:17910:58342#15@0 1177 12 58221643 3 76M * 0 0 GGAGGAGATCAGCGAGAGGCTGGAAGAAGCTGGCGGGGCGACGTCAGCCCAGATTGAGATGAACAAGAAGCGCGAG DDDDDEDDDDDDDDCDDDEEEECEEDFFFGHHJJJJJJJJJIIIJIJIJJJJIIJJIJJIIJIHHHHHFFFFFCCC MD:Z:30C45 RG:Z1126ACXX_7_15 XG:i:0 NH:i:2 HI:i:1 NM:i:1 XM:i:1 XN:i:0 XO:i:0 AS:i:-5 XS:A:+ YT:Z:UU

If I am not mistaken, the first and second reads have the flag 0x8 (i.e. next segment in the template unmapped). Furthermore, the first read has the flag 0x100 (i.e. secondary alignment). From these flags, I understand that these two alignments are the same read aligned in two different places.

After using the FixMateInformation function with this file, these two reads became:

HWI-ST731_145:7:1306:17910:58342#15@0 433 12 57997071 3 76M = 58221643 224573 GGAGGAGATCAGCGAGAGGCTGGAAGAAGCTGGCGGGGCGACGTCAGCCCAGATTGAGATGAACAAGAAGCGCGAG DDDDDEDDDDDDDDCDDDEEEECEEDFFFGHHJJJJJJJJJIIIJIJIJJJJIIJJIJJIIJIHHHHHFFFFFCCC CC:Z:= MD:Z:30C45 RG:Z1126ACXX_7_15 XG:i:0 NH:i:2 HI:i:0 NM:i:1 XM:i:1 XN:i:0 XO:i:0 CP:i:58221643 MQ:i:3 AS:i:-5 XS:A:+ YT:Z:UU
HWI-ST731_145:7:1306:17910:58342#15@0 1201 12 58221643 3 76M = 57997071 -224573 GGAGGAGATCAGCGAGAGGCTGGAAGAAGCTGGCGGGGCGACGTCAGCCCAGATTGAGATGAACAAGAAGCGCGAG DDDDDEDDDDDDDDCDDDEEEECEEDFFFGHHJJJJJJJJJIIIJIJIJJJJIIJJIJJIIJIHHHHHFFFFFCCC MD:Z:30C45 RG:Z1126ACXX_7_15 XG:i:0 NH:i:2 HI:i:1 NM:i:1 XM:i:1 XN:i:0 XO:i:0 MQ:i:3 AS:i:-5 XS:A:+ YT:Z:UU

Now, these two reads does not have the flag 0x8. Furthermore, their RNEXT and PNEXT fields indicate that one is the mate of the other.

This behavior is correct?

Thanks in advance!

[oriol]

If someone using old Picard versions has the same issue: It was a Picard bug. The Picard team solved it very fast in newer versions.

[gmarco]

I'm experiencing weird behavior in cufflinks when using Picard tools FixMateInformation on *.bam produced by tophat2.

Using picard-tools-1.129 and cufflinks v2.2.1.

I'm not getting same results after using fixmate on tophat2 bams. I usually align with tophat2, filter by a quality score of 5 to retain only unique alignments and then fixmate bam with picard tools.

However comparing cufflinks results with and without Picard FixMateInformation make a really big difference. Results with fixmate make no biological sense at all.

Could this be the issue?