Reads missing from alignment

[Jakob Nissen]

Dear Philip

When aligning reads to a short reference, I saw that even small differences between the reference and the read caused a large drop in depth around the area of the divergence. This worries me, because the divergence is much lower that what I would expect would cause such a drop. Specifically, nucleotide identity is still above 85%, and there are stretches of more than 20 bases with 100% identity. BWA MEM finds the matches.

I’ve made a minimal working example, attached. Simply extract the archive and run the shell script.

[Jakob Nissen]

It seems like it works when setting -mrs 0.1. I don’t understand why that is, could it perhaps be better documented what -mrs X does?

• It’s not that the alignment score should be at least X * length(read), since the alignment score for this read pair is around 150 for a 250 bp read.
• It’s not some threshold that sets whether the template should be included, since, in my original non-minimal example, the template was included, but some reads just didn’t map to it.

‌

[ptlcc]

Dear Jakob

Which version of kma are you using, it worked with v1.3.19.

Best,
Philip

[Jakob Nissen]

I was using KMA-1.3.10b. It’s working for me as well with v 1.3.19. Closing this issue

[Jakob Nissen]

Unfortunately, it turned out it only solved the problem for the particular read in the minimal working example. Picking another read, it's still the same.

[Jakob Nissen]

@ptlcc I have updated the minimal example, such that it now also fails for the newest version of KMA

[ptlcc]

Hi Jakob

I have found the bug, but it will take a little more to fix than usual. Fingers crossed en of this week, start of next at the latest.

Best,
Philip

[ptlcc]

Fixed Issue #38

→ <<cset 1f61d5a69d20>>

[ptlcc]

Hi Jakob

It should be fixed now, v1.3.20.

Best,
Philip