Picard MarkDuplicates and Secondary Reads

[Ham.m]

Hi all,

I am running picard MarkDuplicates with REMOVE_DUPLICATES=TRUE.
From what I have read, if picard marks a primary alignment as a duplicate, its secondary alignments should also be marked as a duplicate (and removed). However, I seem to be left with secondary hits after running the above command. For example, when looking at one example read (below), the original file has one primary alignment and one secondary alignment

Original file
-bash-4.2$ grep "1101:14851:94564" Sample006_CA.sam

HWI-ST614:206:C63DHACXX:7:1101:14851:94564 0 MT 16503 60 67M33S * 0 0 GGTTCCTACTTCAGGGTCATAAAGCCTAAATAGCCCACACGTTCCCCTTAAATAAGACATCACGATGGATCACAGGTCTATCACCCTATTAACCACTCAC

HWI-ST614:206:C63DHACXX:7:1101:14851:94564 256 MT 1 60 67H33M * 0 0 GATCACAGGTCTATCACCCTATTAACCACTCAC

(Duplicates removed)
-bash-4.2$ grep "1101:14851:94564" Sample006_rd.sam (after running picard MarkDuplicates)

HWI-ST614:206:C63DHACXX:7:1101:14851:94564 256 MT 1 60 67H33M * 0 0 GATCACAGGTCTATCACCCTATTAACCACTCAC

Can anyone help explain why I am left with secondary alignments? This becomes an issue when running the gatk BQSR as it screens out secondary alignments.

Thank you in advance,
Hannah

[Ham.m]

Found the answer online - http://broadinstitute.github.io/pica...geBamAlignment

If the input is query sorted, rather than coordinate sorted, then the secondary reads are also considered in the duplicates analysis.