Counting and extracting in a new file the unique mapped reads from BWA 0.7.15-r1140 mapping

Question

[valopes]

Hello guys.

I am trying to have a bam file containing just the unique mapped reads from BWA.

I've used to check the mapping results:

samtools flagstat input.bam

and I got

151399675 + 0 in total (QC-passed reads + QC-failed reads)
1299223 + 0 secondary
0 + 0 supplementary
0 + 0 duplicates
150524060 + 0 mapped (99.42% : N/A)
150100452 + 0 paired in sequencing
75050226 + 0 read1
75050226 + 0 read2
141573428 + 0 properly paired (94.32% : N/A)
148964374 + 0 with itself and mate mapped
260463 + 0 singletons (0.17% : N/A)
6491734 + 0 with mate mapped to a different chr
5052530 + 0 with mate mapped to a different chr (mapQ>=5)

I know 150524060 means the unique and multiple mapped reads, but I want to make sure how many unique mapped I have.

So I used:

samtools view -F 0x40 -c input.bam

and I got 75664312

I am not sure if this number above is right.

Could someone please help me with this? Which command should I use to count unique mapped reads? and which to extract in a new bam file just the unique mapped (excluding the multi and unmapped ones)?

Thanks in advance.

[WouterDeCoster]

I added markup to your post for increased readability. You can do this by selecting the text and clicking the 101010 button. When you compose or edit a post that button is in your toolbar, see image below:

[valopes]

Thanks. I didn't know that.

[Kevin Blighe]

Hi, all I know for now is that 0x40 is not the correct flag. Take a look at the SAM specification here: https://samtools.github.io/hts-specs/SAMv1.pdf (go to page 5). I don't believe that BWA records anything specific about unique mappings, but it does retain information about primary (best) and secondary alignments. However, I know that it's possible to infer uniquely mapped reads if you align with Bowtie (bowtie 1), using --best and -m 1 command-line parameters. Bowtie2 doesn't have these parameters but it still retains unique mapping information. When you run Bowtie/Bowtie2, the alignment report gives this information. I've just done some mappings right now (see the part in bold):

5157135 (100.00%) were paired; of these:

805804 (15.63%) aligned concordantly 0 times

2102530 (40.77%) aligned concordantly exactly 1 time

2248801 (43.61%) aligned concordantly >1 times

...

86.78% overall alignment rate

[valopes]

Thank you. As I was used to work with Bowtie2, I was trying to get the same info with BWA. But now I see it is different.

Answer 1

[ATpoint]

Here is a tool that can help you handle the bitwise flags. 0x40 stands for 'first in pair', and samtools view -F stands for discarding those reads flagged as such. Your settings do not really make sense here. If unique means that multimappers are to be excluded and only high-quality alignments are to be counted, I would go with:

samtools view -bh -q 30 -f 3 -F 2316 -c in.bam

It keeps (-f) mapped and properly paired reads that are not multimappers (MAPQ=0), discarding (-F) non-primary alignment, unmapped reads and mates. Play around with the flags to fit your needs.

[valopes]

Thank you. I will try it and let you know how it work for me.