Can gaftools convert an unstable coordinate GAF to a stable coordinate GAF?

[peterdfields]

I'm working on a problem where I'm trying to understand haplotype structure in an admixed mapping population using long-reads mapped to a pangenome genome generated using minigraph-cactus. One intermediate file I'd like to use to begin the analysis is a GAF which includes the stable genome coordinates where a given read maps across the different references included in the pangenome. I've modified the GFA which GraphAligner is using such that every input genome can be treated as a reference and I can get a GAF which includes unstable coordinates as described in https://github.com/lh3/gfatools/blob/master/doc/rGFA.md but I cannot seem to get the stable coordinate GAF. Can gaftools do such a conversion? Thank you for your time and advice.

[samarendra-pani]

Gaftools should be able to do such a conversion. For converting the whole file, please try the following command

gaftools view --gfa <input GFA file> --format stable <input GAF file>

Please note that the code does require the GFA file to have SN, SO, and LN tags.

[peterdfields]

@samarendra-pani Thank you for your response! That's great news that gaftools can create the needed output! My GFA was created with minigraph-cactus -> vg convert with the --vg-algorithm flag in order to use GraphAligner. It does not appear to have SN, SO, and LN tags. Do you have a recommendation for converting my present GFA to the GFA formatted needed to generate the stable GAF conversion?

[samarendra-pani]

I do not know of any tools that can do such a conversion yet. Gaftools assumes the reference graph to have the rGFA format as described here (https://github.com/lh3/gfatools/blob/master/doc/rGFA.md).

Can you post a detailed message on how the GFA you used in your conversion was made? I will recreate the GFA and see if it can be converted into an rGFA (this conversion will also be a good addition to gaftools).

[peterdfields]

Sure! I started with the .gbz file used primarily for aligning short reads with vg giraffe. I was following along with the steps described here for the GraphAligner alternative until vg giraffe releases long-read capabilities. The main command, as described in their example, for creating the GFA is then

vg convert ./hprc10/hprc10.gbz -f --vg-algorithm > ./hprc10/hprc10.gbz.gfa

Let me know if any additional details would be helpful.

*Quick addendum in case it's helpful. Looking at the GFA which minigraph-cactus generates as part of the larger pangenome creation, as compared to the GFA that is created for working with GraphAligner with vg convert, I do see SN, SO, and SR tags but no LN tags. I don't know how this would fit in with gaftools view given this GFA wasn't used as part of the alignment step, ie the GFA created by minigraph-cactus does not work for GraphAligner.

[peterdfields]

Hi @samarendra-pani. I was just wondering if you had had time to have a closer look at this issue? Thank you again for all your help.

[samarendra-pani]

Hello @peterdfields

Sorry for the delay. I am busy with something else now. I looked at the gfa generated by vg convert, and it will be hard to recreate the tags just from the gfa. I will work on a fix next week and let you know.

[peterdfields]

@samarendra-pani That's great news, thank you!

I'm hoping to be able to continue to work within the minigraph-cactus/vg-giraffe ecosystem because it works so well for short-reads (and hopefully soon long-reads). However, might you have a recommendation of another tool chain which could possible generate the requisite inputs (graph, alignment indices, etc.) so that gaftools could generate the stable coordinate GAF?

[peterdfields]

@samarendra-pani I just noticed that vg giraffe now offers long read mapping: https://github.com/vgteam/vg/releases/tag/v1.63.0

Do you think this could simplify the problem at all for creating the stable coordinate GAF?

[samarendra-pani]

I can try mapping with vg giraffe later. I am in the middle of writing a script for doing the conversion. Need to make sure it works properly with some test cases.

Does vg giraffe give an option to output a file in stable coordinates? If that can solve the problem, then awesome.