maickrau
unspecified edge overlaps error
I'm running GraphAligner bioconda 1.0.16- with a gfa that contains "*" edge overlaps. GraphAligner is failing with the following error.
Error in the graph: Unspecified edge overlaps (*) are not supported
I saw the release notes for v1.0.13 state Allow graphs where edge overlaps are "*", assume that means zero overlap. Is this functionality no longer supported in v1.0.16?
This functionality is no longer supported as of v1.0.15 and later. The reason is that sometimes assemblers might output * edge overlaps when there is an overlap. GraphAligner now requires all edge overlaps to be explicitly specified.
Thanks for the info, I understand why you wouldn't want to assume "" = 0 overlap. Have you considered adding an option to explicitly set the "" overlap value? I have several large graphs with no edge overlaps. Currently, I would need to maintain separate, modified graphs just for GraphAligner.
There's currently no plan to add an option for that. Do the other tools you use require * overlaps? If not, my suggestion would be to replace the overlaps with 0M and use those graphs.
I'm primarily working with vg, which was used to produce the GFAs in question. Of course, the * overlap GFAs produced by vg work fine with all vg tools. I had modified the link/edge overlaps to use '0M' instead of * for GA. I have confirmed that vg is still able to build identical index files using both the original and '0M' modified GFA. It seems I can safely proceed with using '0M' modified GFAs for both tools.
I suppose not many people are using vg to construct graphs. I was the one who submitted the request for vg * overlap handling over a year ago. (https://github.com/maickrau/GraphAligner/issues/37) After a couple of GA releases with dropped support for vg-based GFAs, I haven't seen any mention of the issue.
I am noticing some other issues using the GA gam (using '0M' GFA) with some vg tools, but I'll create a separate issue for that.