why those reads don't contain 0x2

Question

[Tao]

I find 0x2 means read mapped in proper pair.As shown in the following image

And why my reads(following image) don't contain 0x2?

[Carlo Yague]

The exact definition of "mapped in proper pair" depends on the aligner and the options used for mapping. What did you use ?

[Tao]

use bwa mem and no else options

[Tao]

I want to know why those reads don't contain 0x2,Not my bam file all don't contain 0x2

Answer 1

[Carlo Yague]

The exact definition of "mapped in proper pair" depends on the aligner and the options used for mapping. In bwa mem, the algorithm automatically infers the general read orientation and the insert size distribution from a sub-sample of reads. It should have printed it in the standard error output when you ran the program. Pairs with different read orientation or with insert size too high or too low compared to the averaged insert size distribution are considered not mapped in a proper pair.

[Tao]

In red line marking reads,I think it's not too high or too low(because I see some reads insert size is higher than it,also have some lower),And read1 mapped in Watson strand read2 mapped in Crick strand.its read orientation and the insert size distribution is all right,But it don't contain 0x2

[i.sudbery]

Difficult to tell without seeing the original output from mem with the distribution. I think its also possible that mem will mark reads not proper pair if they are too close, as well as too far.

To be honest, I rarely use the proper pair flag as its so ill defined. If I want to filter reads, I deduce my own rules, and then filter the reads manually.

[Carlo Yague]

Then it is very weird if you say that the numbers are within the range of the population. I don't see anything wrong with the orientation or insert size of those reads. Perhaps the file got corrupted somehow. You could try to use samtools fixmate and see if it fixes the issue.

[Tao]

Thanks a million！