MarkDuplicates in picard

[bair]

Hello all,

What's the metrics file output from Markduplicates function in picard?

Can I get how many reads marked as duplicates in this file?


Thanks

[mard]

Yes it tells you the number of reads that have been marked as duplicates, as well as the total number of reads. But note that reads that Picard marks as duplicates do not necessarily have identical sequence they just map to the same chromosomal location.

[bair]

Yes it tells you the number of reads that have been marked as duplicates, as well as the total number of reads. But note that reads that Picard marks as duplicates do not necessarily have identical sequence they just map to the same chromosomal location.

Thanks. How to pick up the duplicates to remove? keep the best alignment one if they do not have identical sequences?

Here is what I got from picard :


## METRICS CLASS net.sf.picard.sam.DuplicationMetrics
LIBRARY UNPAIRED_READS_EXAMINED READ_PAIRS_EXAMINED UNMAPPED_READS UNPAIRED_READ_DUPLICATES READ_PAIR_DUPLICATES READ_
PAIR_OPTICAL_DUPLICATES PERCENT_DUPLICATION ESTIMATED_LIBRARY_SIZE
Unknown Library 27221401 548559917 190908169 14563968 58165860 0 0.11642 2400441897

## HISTOGRAM java.lang.Double
BIN VALUE
1.0 1
2.0 1.795707
3.0 2.428856
4.0 2.932657
5.0 3.333535
6.0 3.652516
7.0 3.906332
8.0 4.108295

What is this histogram about?

My original bam file has 657624702 paired reads, so 2*657624702 in total. After remove duplicates, bam file has 1184353716 reads in total. So suppose,
2*657624702 - 1184353716 = 130895688 reads removed.

I couldn't get this number from picard output M file, any help?

Thanks

[psm3426]

The reason for the histogram is one of the FAQ on their wiki.

http://sourceforge.net/apps/mediawiki/picard/index.php?title=Main_Page#Q:_What_is_meaning_of_the_histogram_produced_by_MarkDuplicates.3F

The reason that you couldn't get that number is because for read pair duplicates, they divide the actual number of duplicates in half before reporting it. So in your case, 2 * 58165860 (value under paired_read_duplicates) = 130895688, which was the number of duplicates you were missing. =)