Question on duplicate removal with STAR

[graceqy]

I was trying to do duplicate removal after alignment with STAR, and here is how i ran the command line:

$ STAR --runMode inputAlignmentsFromBAM --inputBAMfile /dir/ti/my.sorted.bam --bamRemoveDuplicatesType UniqueIdentical

I expected the resulting bam files to be smaller than the input. However, the output bam files are bigger than the input one. Is this reasonable? Any explanation or reason that caused this?

Thanks.

[bowhan]

1. with --bamRemoveDuplicatesType UniqueIdentical, STAR does not "remove" the duplicates from the bam, but mask them as PCR duplicates by turning on a bit in the bitflag field. If you do `samtools view -c` on the input and output files, they will have exactly the same reads. But if you do `samtools flagstat` on them, the output will have PCR duplicates information. Now you can use `samtools view -F0x400` to remove them.
2. the reason for the increased size is because the output is NOT sorted. Sorted bam is usually smaller than unsorted ones due to better compression efficiency. After you sort the output bam, they should have very similar sizes.

[graceqy]

This really helps. Thank you very much!