New workflow for hal2vg

[RenzoTale88]

Hello,
I'm tring to convert a branch of a hal archive by specifying the root genome as follow:

hal2vg --noAncestors --chop 32 --hdf5InMemory --rootGenome Anc04 --progress out.hal > VG/out.raw.pg

Once I get this archive, I try to use clip-vg to forwardize the reference genome:

clip-vg -u 100000 -e REFERENCE --progress VG/out.raw.pg > VG/out.clip.pg

When I run the second comment, I get the following error:

[clip-vg]: Loaded graph
[clip-vg]: Loaded 148 BED intervals over 146 sequences
[clip-vg]: Clipping 1 intervals from path GENOME1.seq2.1
[clip-vg]: Clipping 1 intervals from path GENOME1.seq2.1
....
[OTHER sequences here]
...
[clip-vg]: Clipping 1 intervals from path GENOME1.seq10.1
[clip-vg]: Clipping 1 intervals from path GENOME1.seq11.1
[clip-vg]: Clipping 1 intervals from path GENOME1.seq12.1
[clip-vg]: Clipped 25502192 bases from 809839 nodes in 146 paths
[clip-vg]: 114514 orphaned connected components were removed with total 390836444 bases
[clip-vg]: Forwardized 4665 / 7468 steps in reference path REF.NW_XXX.1
[clip-vg]: Forwardized 2139 / 2218 steps in reference path REF.NW_XXY.1
[clip-vg]: Forwardized 1732 / 4106 steps in reference path REF.NW_XXZ.1
[clip-vg] error: Cycle detected in reference path REF.NW_XYZ.1 at node 99701782

I'm probably missing some trivial step to break the cycles in the alignments, but I'm unsure what.
Thank you for the help!
Andrea

[glennhickey]

clip-vg can't (in general) forwardize a cyclic reference path and for that reason it won't even try. For example there is no way to forwardize a path that looks like

>1 >2 <1

The minigraph-cactus pangenome pipeline guarantees an acyclic, forward reference path, but regular cactus does not.

[RenzoTale88]

@glennhickey thanks for clarifying this. Is there a comparable way to extract a subgraph from a hal archive generated with the standard cactus workflow? Something comparable to the old --refGenome option.

[glennhickey]

Did the old --refGenome pull apart duplications? Maybe it did. I guess you could imitate that by running halRemoveDupes before -clip-vg. Without doing that, I can't see how one could (for reasons above) forwardize a standard hal.

[RenzoTale88]

Ok thanks for this, I'll give it a go!

[RenzoTale88]

@glennhickey unfortunately, adding halRemoveDupes before running hal2vg and clip-vg didn't fix the issue. I'll try using it without the clip-vg step. Thanks anyway for the help!

[glennhickey]

Hmm, that's unexpected. I'd be happy to take a look if you want to share the data.

The only real consequence of not having a "forwardized" reference path is that it can't be represented in rGFA tags, which I doubt is a concern for you.

Still, if you're aligning distant species with cactus, I'd be concerned in general of the output vg graph being to complex to do anything with..

[RenzoTale88]

@glennhickey thanks for the offer, but I do not need to perform any alignment with VG and/or conversion to rGFA. I am simply using the PG to identify all nodes not coming from one genome, and assign them to the other used. I'm expecting them to work, but in case I come across issues with it I'll come back to you for help. Thanks again!