VG : No reference-sense paths available in the graph; falling back to generic paths.

Question

[Hang]

Hi,

I was aligning short reads to a customized gfa file using vg giraffe. I tried to extract mapped reads from the gam files, first using vg surject to convert the gam file to a bam file.

However, I had this error "No reference-sense paths available in the graph; falling back to generic paths." I tried to add a reference path line in my gfa file like "P ref_path 1+,2+,3+ ,,*". However, the error still remains.

Is there a way to add a reference path in the gfa file?

Thanks in advance,

Hang

Answer 1

[anovak]

If your GFA has paths in it that are P lines with names that don't include a sample name, contig name, and separators, then those are what vg thinks of as "generic" paths, and vg surject will surject into them. It will print the message about falling back to generic paths in this case; this is not an error and vg surject is able to continue, assuming there are actually generic paths in there that it can surject to.

If vg surject finds any paths that it thinks look like a primary reference, it will prefer those paths and not use any generic paths. Reference paths need to be associated with a sample, a phase number, and a contig: they should have P line names that follow PanSN format like GRCh38#0#chr1 or should be W lines with the relevant fields filled in. Additionally, the sample name needs to be marked as a reference sample using an RS tag on the GFA header line, so that vg can tell that GRCh38 is a reference assembly and not an individual.

vg surject will ignore any "haplotype" paths, which are paths associated with a sample that is not marked as being a reference assembly.

You can use vg paths --metadata -x graph.gfa to see how vg interprets the paths it sees in a graph under its path metadata model.

If you have a Giraffe GBZ file instead of a GFA, you can set the sample names that are marked as reference samples by following https://github.com/vgteam/vg/wiki/Changing-References