how to get reads with "mate mapped to a different chr"

[gene_x]

In the output of

samtools flagstat input.bam

There is these last two lines:

xxx +0 with mate mapped to a different chr
xxx +0 with mate mapped to a different chr (mapQ>=5)

I'm wondering what's the FLAG for "mate mapped to a different chr"?

[Brian Bushnell]

There is no flag; you compare "rname" with "rnext". If they are the same, or "rnext" is "=", and they are both mapped (0x1 set, 0x4 and 0x8 not set) then they are on the same sequence. They are mapped to different sequences if both are mapped but rnext is not "=" and not equal to rname.

[gene_x]

There is no flag; you compare "rname" with "rnext". If they are the same, or "rnext" is "=", and they are both mapped (0x1 set, 0x4 and 0x8 not set) then they are on the same sequence.

oh, i see. Thanks!

[Richard Finney]

Good question.

Consult the Sam format documentation.

https://samtools.github.io/hts-specs/SAMv1.pdf

Section 1.4 contains the bitwise flag descriptions.

Note that not all software properly sets these flags.

I guess you'd check the 0x4 and 0x8 flags ( 0x4=segment unmapped, 0x8=segment unmapped). If both unmapped then check if field 3 (RNAME) is not same as field 7 (RNEXT) [and 8 field is not '*' and not '='] .

There are various "FIXMATE" programs running around; you may wish to use.

I'm not sure about 0x800 flag (not same as 0x8). "Chimeric" I'm guessing doesn't always mean same chromosome.

[Brian Bushnell]

I'm not sure about 0x800 flag (not same as 0x8). "Chimeric" i'm guessing doesn't always mean same chromosome.

I don't think the answer is fully defined for either chimeric or secondary alignments.

[gene_x]

I don't think the answer is fully defined for either chimeric or secondary alignments.

I'm actually pretty confused about what is "secondary alignment".. can you clarify it a little bit?

Also, in the attached image, I don't understand how the entry in the second column equals to that in the first column? For example, 0x0001 and p are supposed to represent the same thing I guess? what's the encoding conversion rules here? And where is the string representation used? In BAM files?

[Brian Bushnell]

I have not seen that string notation before ("p", "P", etc) and it's not part of the SAM specification as far as I know. Reads are supposed to have at most one primary alignment; if a read maps to multiple locations, it can have multiple secondary alignments (0x100 flag bit). But reads can also have multiple 0x800 "supplementary" alignments, a new feature of the sam format which is rather confusing.

[gene_x]

I have not seen that string notation before ("p", "P", etc) and it's not part of the SAM specification as far as I know. Reads are supposed to have at most one primary alignment; if a read maps to multiple locations, it can have multiple secondary alignments (0x100 flag bit). But reads can also have multiple 0x800 "supplementary" alignments, a new feature of the sam format which is rather confusing.

Yeah, the documentation is very limited and confusing. I have to search online for a few tutorials to get an OK understanding of bitwise FLAG..

I'm confused.. what does supplementary alignment mean?

[Brian Bushnell]

In practice, if you run bwa, it means a chimeric alignment, in which there are multiple local alignments that are not very close to each other. I am not aware of any other aligners that generate it.

[gene_x]

In practice, if you run bwa, it means a chimeric alignment, in which there are multiple local alignments that are not very close to each other. I am not aware of any other aligners that generate it.

I see. Thanks.

[dpryan]

I have not seen that string notation before ("p", "P", etc) and it's not part of the SAM specification as far as I know.

It's been deprecated. Back in the day, there was a multicharacter version of the FLAG field available from samtools. That was done away with after the conversion to htslib (I'm pretty sure it was still there in 0.1.19). The only thing I've ever seen use those is BS-Seeker2, in fact.

[gene_x]

It's been deprecated. Back in the day, there was a multicharacter version of the FLAG field available from samtools. That was done away with after the conversion to htslib (I'm pretty sure it was still there in 0.1.19). The only thing I've ever seen use those is BS-Seeker2, in fact.

I see. Good to know!

[splaisan]

one way to get it

see https://www.biostars.org/p/17575/#327644