Bowtie2 YT tag question

[Aizon]

Hello!
I am currently working with a dataset of Illumina PE reads (length 100bp, fragment 500bp). Now, I know the definition for Bowtie's YT tag and the various values "CP", "UU", "UP", "DP", but I have stumbled upon an example where I do not understand why the pair has been assigned the YT:Z:UP tag.

Here are the pairs alignment:

FCC0WRYACXX:4:1115:3531:40304#ATGAGGAA 113 Chr6 3218530 12 100M Chr2 3121587 0 ATATCGCA...AGCTATCA qual AS:i:-5 XS:i:-17 XN:i:0 XM:i:1 XO:i:0 XG:i:0 NM:i:1 MD:Z:9T90 YT:Z:UP

FCC0WRYACXX:4:1115:3531:40304#ATGAGGAA 177 Chr2 3121587 0 100M Chr6 3218530 0 ACTACTAG...ATCGATAG qual AS:i:0 XN:i:0 XM:i:0 XO:i:0 XG:i:0 NM:i:0 MD:Z:100 YT:Z:UP

I would expect these to be considered discordantly mapped (DP) and not UP. Admittedly one of them has a MAPQ of 0, but it is mapped. Should have an '*' in some fields if it was not mapped (and then I would understand the UP.

Any ideas how to explain this behaviour?

I tried finding the similar problem mentioned in other threads but couldn't find quite the same situation.
Cheers

[dpryan]

N.B., ignore my earlier reply, it was wrong (thus I deleted it).

If the reads align to different chromosomes then they're no longer considered discordant, which would normally apply if the insert size was just too big.

[Aizon]

Thank you for your reply dpryan!

However, I'm pretty sure mates that have mapped on different chromosomes are tagged with YT:Z: DP
I cannot check right now, but I will reply again tomorrow with an example.

As promised, reads that map to different chromosomes are still considered (and I think rightfully so) Discordant Pairs by bowtie. Here's an example, from the same output as the example in my first post:

FCC0WRYACXX:4:1104:2072:41391#ATGAGGAA 81 Chr1 8014059 42 100M Chr2 4225437 0 CTACTA... C>>DDC... AS:i:0 XN:i:0 XM:i:0 XO:i:0 XG:i:0 NM:i:0 MD:Z:100 YS:i:0 YT:Z: DP

FCC0WRYACXX:4:1104:2072:41391#ATGAGGAA 161 Chr2 4225437 42 100M Chr1 8014059 0 GTTGCA... B@BFFF... AS:i:0 XN:i:0 XM:i:0 XO:i:0 XG:i:0 NM:i:0 MD:Z:100 YS:i:0 YT:Z: DP

[dpryan]

Oh the many mysteries of bowtie2. I'm guessing that this behavior is described by this comment in the source code (for context, this is from aln_sink.cpp and both concordant and discordant alignments have already been dealt with):

// If we're at this point, at least one mate failed to align.
// BTL: That's not true. It could be that there are no concordant
// alignments but both mates have unpaired alignments, with one of
// the mates having more than one.

This still leaves some ambiguity. I could read this to mean that you'll get YT:Z: DP unless there's more than one valid secondary alignment or that you'll get YT:Z: DP unless there's a secondary alignment of equal score to the primary. If you have a chance, look for more examples in your alignments where the mates are on different chromosomes and they have YT:Z: DP. If you tracked their MAPQ scores, then that'd likely clarify this (if they're all 42, then the former reading is likely correct, if not, then presumably the latter is correct).

[Aizon]

This is getting confusing

So if I understand this correctly, there shouldn't be pairs with mate mapping on different chromosomes, tagged DP, and MAPQ 0. Right?

[dpryan]

I'd look for MAPQ 1, actually. You can get a MAPQ of 0 if there's a valid secondary alignment that's still not as good as the primary (yes, this seems backwards).

[Aizon]

There are no alignments where I have DP, different chromosomes and MAPQ 1.

However, there are cases where I have DP, different chromosomes and MAPQ 0 for one or both mates, as well as cases where I have UP, different chromosomes and MAPQ 0 for one or both mates.

Just as a note, in general my output is fairly normal and I have worked already for quite some time on it. But I just noticed that there are some cases where I don't understand how bowtie2 worked exactly.

PS: Thank you so much for looking into this!

[dpryan]

That makes sense then.

0 is a weird case and there are 2 ways that you can get it. The ones leading to UP are the case where the AS:i: and XS:i: tags are the same (i.e., the best and second best alignments are identical), or at least that'd be my expectation. Whether these alignments have MAPQ 0 or 1 depends on how high/low the AS:i: value is and what --score-min is set to. The cases where MAPQ 0 alignments have UP are likely when AS:i: is greater than XS:i:, but AS:i: is still relatively low (the actual algorithm is a bit weird).