Poor variation granularity in graph creation from assemblies with minigraph-cactus

[Mirkocoggi]

I'm trying to create a pangenome graph with minigraph-cactus from multiple assemblies (fasta) but I'm encountering a problem in the granularity of the variations identified.

In particular, sometimes big super bubbles are created with very similar alternative sequence nodes, instead of collapsing the common regions among them. I provide an example:

S 1 ACCGCTCGCGCGTTAC
S 2 ACCGCACGCGCGTTAC
S 3 ACCGCACGCGCGATAC

The three nodes are alternatives belonging to the same superbubble, but it is clear that the only differences are SNPs: T instead of A at pos 6 and A instead of T in position 13. It seems like some small variations are not identified and it chooses to create big bubbles of long alternative sequences.

I don't know if it could have helped but I tried to use the --collapse flag, which unfortunately didn't work (It crashed in multiple combination of settings).

[glennhickey]

I've seen under-alignments in large, very complex regions. --mgSplit can help a bit with this. At the smaller scale in your example, I'm not sure what the cause would be. Are you sure that one node isn't an insertion with respect to the others? What is the bubble exactly?

[glennhickey]

Yeah, --mgSplit won't help. --collapse may help, but not all types of indexing will work with it.

[Mirkocoggi]

Hi, the bubble I wrote was only an example to explain the problem. Here I'm appending one of the bubbles I found in the graph after appling Minigraph-castus. In this snapshot all the sequences are similar but the tool does not recognize the similarity.

[glennhickey]

That does seem pretty bad. And those nodes are all paths through the same bubble?