Adding Unpaired_Reads_Examined And Read_Pairs_Examined In Metrics Of Picard Markduplicates

Question

[ccshao]

Adding the number of UNPAIRED_READS_EXAMINED and READ_PAIRS_EXAMINED does NOT give me the number of original file.

Here is my data: original file has 91137584, they are reliable reads by the command "samtools view -S -bq 1".

I want to remove the duplicate with picard, and hence set the "REMOVE_DUPLICATES=true". Output of metrics is:

UNPAIRED_READS_EXAMINED    READ_PAIRS_EXAMINED    UNMAPPED_READS    UNPAIRED_READ_DUPLICATES    READ_PAIR_DUPLICATES    READ_PAIR_OPTICAL_DUPLICATES    PERCENT_DUPLICATION    ESTIMATED_LIBRARY_SIZE
104579    45330048    0    68012    14161407    5034139    0.312796    74939630

the duplicate marked file has 62746758, and 91137584 - 14161407*2 - 68012 give the number of reads in new bam file.

However, 104579 + 45330048*2 gave me 90764675, less than the actual number of reads in my file, 372909 reads are missing.

Could someone explain me what is going on?

Thanks!