count the uniquely and duplicates reads for BWA bam file

[mrfox]

Hi all,

If this topic has been solved in a previous post, I apologize for that. And please direct me to the solution.

I used BWA to align exome sequencing data. And I want to count the number of uniquely mapped reads and the number of duplicates reads (not multi-hits reads) in the obtained BAM files. Samtools flagstat is supposed to work: uniquely = mapped - duplicate. Unfortunately samtools flagstat always get "0+0 duplicate" for BWA aligned BAM files.

I wonder if anybody has an easy solution for this.

Thank you all.

[swbarnes2]

Hi all,

If this topic has been solved in a previous post, I apologize for that. And please direct me to the solution.

I used BWA to align exome sequencing data. And I want to count the number of uniquely mapped reads and the number of duplicates reads (not multi-hits reads) in the obtained BAM files. Samtools flagstat is supposed to work: uniquely = mapped - duplicate. Unfortunately samtools flagstat always get "0+0 duplicate" for BWA aligned BAM files.

I wonder if anybody has an easy solution for this.

Thank you all.

Samtools flagstat is just reading the flags in the bam lines. If you haven't used software that will assign the "duplicate" flag in your bam, flagstat won't know that you have duplicates.

Software like Picard's MarkDuplicates will flag reads as duplicates where appropriate. samtools rmdup will just get rid of duplicates.

[mrfox]

Hi swbarnes2,

Thanks for your reply. I seldom use Picard's MarkDuplicates but it seems to me that if we apply it on a BAM file and then the duplicates will be labeled and then we are able to count the number of duplicates in the new BAM. I will try that. Please let me know if you have a better idea.

Thank you.

[swbarnes2]

Hi swbarnes2,

Thanks for your reply. I seldom use Picard's MarkDuplicates but it seems to me that if we apply it on a BAM file and then the duplicates will be labeled and then we are able to count the number of duplicates in the new BAM. I will try that. Please let me know if you have a better idea.

Thank you.

Yeah, that's fine. Or use samtools rmdup, and count how many reads are missing from the de-dupped file.

Those tools should yield almost exactly the same thing (one known exception, rmdup won't touch reads where each read falls in a different chromosome, MarkDuplicates should deal with those properly.)