CIGAR should have zero elements for unmapped read

[doc2r]

Hi Everyone

I am running MarkDuplicates.jar on my paired end-bwa-mapped BAM file. However I get a weird error message, in fact there are 1000’s of these actually (0.5M so far).

An example is

Ignoring SAM validation error: ERROR: Record 68, Read name MachineName:1:2:15520:114537#0, CIGAR should have zero elements for unmapped read.

Has anyone experienced this before?

When I pull out the reads from the bam file above I see the read details as follows
(read1)
MachineName1:2:15520:114537#0 73 chr10 50281 0 69M31S = 50281 0 CTGTGCAATAACTGTGTACAAAAGCCCCAAAGCTTAAATTGTGCAGTTGAGCGCATGTTCTGTTGTTCAGCATTTATGTTGGTTTATAGTGGAAAAGATT
?5<3;2><<62@3A<<<7>@@=B7BCC=BB:,<+:9/)<+0;*'+-'271B@BB2BC@CC=B0B<>BA################################
XC:i:69 XT:A:R NM:i:3 SM:i:0 AM:i:0 X0:i:2 X1:i:0 XM:i:3 XO:i:0 XG:i:0 MD:Z:8A24A3T31


(Read2)
MachineName:1:2:15520:114537#0 133 chr10 50281 0 88M12S = 50281 0 TTCATTGTTTGGCATAACAGTACTTCAGATTTGAATCATCTAATAACATTGTCATCATAGCATATTCTCCTGGAAGTAACACACAATAACTACTTCAAAA
E/EBEDDFDFEFFF=?@CBA.=>.<.9.::EE=EE33?<7D>BCB-<5<>.:37:@<B/<.8986<:9;->A@A@BADBB@BB@AEE############# XC:i:88


Oddly enough they map to the same position……Although the sequences are completely different. I BLAT’ed the sequences and found for read one and two respectively

SCORE START END QSIZE IDENTITY CHRO STRAND START END SPAN
88 1 100 100 94.0% 10 + 50281 50380 100
86 1 100 100 93.0% 10 - 50520 50619 100


So sequence two really maps to a different position on chromosome 10 at a distance that’s roughly the expected insert size….

Could it be because the %Identity is low? that BWA mapped the pair incorrectly?

[swbarnes2]

Read two didn't map at all. bwa gives it the mapping coordinates of the read that did map. It's so that they will sort together by coordinate. You can tell from the flag: 133 = 128+4+1 = second read of pair + unmapped + paired read.

The flag for read one is 73 = 64+8+1 = first read of a pair + mate didn't map + paired read

I'm not quite sure what Picard does with that in markduplicates. If you use stringency Lenient, it won't yell at you about the CIGAR in unmapped read, but I don't know if it will mark duplicates properly.

At the end, you can filter your .bam for read that don't have a 4 in the binary flag, that'll get rid of all the unmapped reads.

[doc2r]

Thanks swbarnes2,
Appreciate the response and the lesson.
This was most helpful