Minimap2: MD tag

[fritzsedlazeck]

Hi,
are you planing to provide the MD tag at some point? Or should I rely on samtools to reconstruct this information?

Thanks
Fritz

[lh3]

My current plan is to ditch MD and introduce a new cs tag instead. The problem with MD is that it is not a standalone tag. You have to parse query sequence, cigar and MD at the same time. In comparison, cs encodes cigar. You can parse cs alone.

[fritzsedlazeck]

Interesting, would make life easier for sure.
Thanks
Fritz

[alexiswl]

Hi Fritz,
Working with Pysam I also found this particularly frustrating,
methods such as get_reference_sequence() rely on the MD tag, I'm sure it's only a matter of time before they can incorporate the cs tag as a substitute.

I found that running
samtools calmd bam_file reference_file > new_bam_file would add in the md tag as needed.

Cheers,
Alexis.

[fritzsedlazeck]

Hi Alexis,
This is certainly true but very slow. Minimap2 has all the information during the running time since it writes it out in the cs string.

I will incorporate the cs string soon into my program to support the necessary step to replace cigar and MD string with it.

Cheers
Fritz

[lh3]

Oh, I forgot that you had requested the MD tag, @fritzsedlazeck. There have been a few independent requests on MD recently, so I have decided to implement it in minimap2. The next release will support MD.

[fritzsedlazeck]

Are you still pushing for the cs string? I think it would make things easier for us compared to MD + cigar...

[lh3]

cs will certainly stay in minimap2. The PAF output needs this tag anyway as PAF doesn't store the query sequence like SAM. My assembly-to-assembly variant caller is unable to work MD, either.

[gevro]

I think it would be nice to add a tool to samtools to generate the cs tag, similar to samtools calmd.