advice for visualizing a graph genome constructed by "vg construct"

Question

[Jong]

Dear vg team and users,

I have constructed a graph genome using one linear reference genome and SNPs & SVs called based on the genome. The code context used is: vg construct -a -f -S -r ${REFERENCE}.sorted.fa -v ${SNPs_SVs}.normed.vcf.gz -m 1000000 > ${GRAPH}.vg

So, this vg file does not have non-refernce contigs and haplotype information (although it has alt information), and so it seems SequenceTubeMap can't display its structure. GrahViz's dot and "vg viz" also could not display due to variants of large sequence length. (i.e., "dot" spit out an error regarding text size limit it can handle; "vg viz" generates 48Mb image file that image viewers cannot handle and even when I saw briefly, no variants were shown.)

In this case, could anyone advise me which approach I should use to visualize a graph genome of this structure?

Thanks for your advice in advance.

Answer 1

[Jordan M Eizenga]

Bandage-NG can work for somewhat larger graphs than dot, although the layouts can sometimes be a bit deceptive. You might also be interested in odgi viz or odgi layout+odgi draw.

If you're trying to visualize a graph with 1,000,000+ bp through, you might have a hard time finding a tool that can simultaneously show you the fine-scale and large-scale features of the graph. You might have a better time setting -m to a smaller value, extracting regions of interest (with vg chunk), and then using vg view+dot or one of the other tools.

[Jong]

Appreciate your answer. Unfortunately, Bandage-NG and odgi both are not working for this graph genome either... I will need to try more search to find a tool. Thanks, though!

[Jordan M Eizenga]

It's hard to be sure without more elaboration on what you mean by "not working", but it might be a question of file formats. Both Bandage and odgi will expect GFA format, which you can convert to using vg convert.

[Jong]

Thanks for your answer. I guess the problem is that this graph genome, constructed using one reference and a vcf file, does not have assemblies/contigs (and so not haplotypes) weaved onto the reference. So odgi, for example, just shows the reference as a single bar, but nothing else. I have tried ‘vg view’ to convert the problematic graph genome into gfa file, but no avail yet. In contrast, another graph genome, constructed through minigraph-cactus using a reference and multiple assemblies, could be successfully visualized by odgi. Maybe I need to use ‘vg convert’ rather than ‘vg view’? I will try that route. Thanks!

[Jordan M Eizenga]

Bandage-NG doesn't use haplotype paths at all, so it should be unaffected by that. I'm not positive about this, but I believe that odgi will still show the edge structure of the graph even without haplotypes (as rectangle-shaped arcs). However, it's possible that, in a VCF graph, the edges don't reach very far along the reference, and so they might be too small to be visible if you're looking at the whole graph. This is the issue that I warned about earlier with the fine-scale and large-scale features.

[Jong]

Im visualizing just a gene so variants should be fine-scale features. It seems I should try odgi 1D to try to see the arcs or Bandage-NG with gfa files. Thanks very much for the answers. This helped me a lot!