vg surject for long reads

Question

[ved_vyas]

Can vg surject be used for long reads like ONT or pacbio ?

[Jordan M Eizenga]

I can't say that we've extensively characterized the performance with long reads, but in principle yes, it should work. As of right now, VG itself doesn't have a mapping tool that works well for long reads (although we probably will in the future). However, you could perhaps use alignments from GraphAligner. One speed bump we've discovered in the past is that GraphAligner's GAF output isn't compatible with vg surject, so if you go that direction you should use GAM as the alignment format.

[ved_vyas]

Thank you for the prompt reply, I did exactly what you have described, but got the following error:

warning[vg::Surjector]: Refusing to perform very large alignment against 119596 bp strand split subgraph for read Read_name ; suppressing further warnings. terminate called after throwing an instance of 'std::out_of_range' what(): basic_string::substr: __pos (which is 22) > this->size() (which is 1)

Originally I had the gfa file, which I converted to vg and xg index using "vg convert" module: In GraphAligner vg index was used for mapping and in vg surject xg index was used. Command used for gam to bam conversion: vg surject -x ${xg_file} -p chr21 -t 20 -b ${gam_file} > ${bam_file}