Using fragmented assemblies to build a graph

[panosioannidis]

I have a human pangenome graph (rGFA format) that I built with minigraph. All 24 genomes used for this graph were very contiguous (PacBio HiFi with N50 > 10Mbp). I used CHM13 as the backbone.

So, I augmented this graph with another 119 assemblies that I obtained from short read sequencing. But since these were fragmented I decided to only keep contigs >10Kbp (I got some 10,000 to 20,000 per assembly). I ran minigraph like this:

minigraph -l 10000 -d 10000 -cxggs -t16 out.gfa $(ls Th_*.fa.gz) > out_aug.gfa

where out.gfa is the rGFA graph from the high quality assemblies and the Th_*.fa.gz are the filtered, short read assemblies.

The weird thing is that in the output graph (out_aug.gfa), some of the short read samples are not found at all. For example, even though one of the input assemblies was Th_003.1.fa.gz, when I grep Th_003 out_aug.gfa I get nothing! Looking at the minigraph log, I found:

...
[M::ggen_map::512.128*2.41] loaded file "Th_003.1.fa.gz"
[M::ggen_map::521.605*2.66] mapped 24362 sequence(s) to the graph
[M::mg_ggsimple_cigar::523.034*2.65] inserted 4 events, including 0 inversions
[M::mg_index::628.549*2.47] indexed the graph
[M::mg_opt_update::633.564*2.46] occ_max1=100; lc_max_occ=2
...

This means that the file was processed, right? If it was why isn't it found in the output graph?

[lh3]

Your contigs need to be named with Th_003 if you want to grep them.

[panosioannidis]

Thanks for the reply Heng!

The names of the contigs do contain Th_003, if that's what you mean (e.g. >Th_003#4#contig_9028260). It's not just the file name that contains Th_003.