Question regarding -L (min variant length) parameter

[cjain7]

Ideally, -L 1 should include maximum number of variants in genome graph, but this does not happen. Please see the runs below:

$ minigraph -cxggs -L 2 hap1.chr18.fa hap2.chr18.fa > output.gfa
[M::main::0.286*0.55] loaded the graph from "hap1.chr18.fa"
[M::mg_index::2.271*1.22] indexed the graph
[M::mg_opt_update::2.422*1.20] occ_max1=66; lc_max_occ=2
[M::ggen_map::2.900*1.07] loaded file "hap2.chr18.fa"
[M::ggen_map::24.119*1.54] mapped 2 sequence(s) to the graph
[M::mg_ggsimple_cigar::24.243*1.54] inserted 5736 events, including 67 inversions
[M::main] Version: 0.17-r529-dirty
[M::main] CMD: minigraph -cxggs -L 2 hap1.chr18.fa hap2.chr18.fa
[M::main] Real time: 24.353 sec; CPU: 37.454 sec; Peak RSS: 1.531 GB

$ minigraph -cxggs -L 1 hap1.chr18.fa hap2.chr18.fa > output.gfa
[M::main::0.286*0.55] loaded the graph from "hap1.chr18.fa"
[M::mg_index::2.271*1.22] indexed the graph
[M::mg_opt_update::2.422*1.20] occ_max1=66; lc_max_occ=2
[M::ggen_map::2.900*1.07] loaded file "hap2.chr18.fa"
[M::ggen_map::24.119*1.54] mapped 2 sequence(s) to the graph
[M::mg_ggsimple_cigar::24.243*1.54] inserted 0 events, including 0 inversions
[M::main] Version: 0.17-r529-dirty
[M::main] CMD: minigraph -cxggs -L 2 hap1.chr18.fa hap2.chr18.fa
[M::main] Real time: 24.353 sec; CPU: 37.454 sec; Peak RSS: 1.531 GB

I understand that -L 1 is far below the intended default value (50), but the above behaviour appears to be weird. Minigraph when run with -L 1 recognises 0 variants in the sequences. I was curious to visualise the graph that includes all het variants. This appears to be a bug.

Thanks!

[ASLeonard]

Did you test other values between 2 and 50 @cjain7 ? Curious if you observe a general increase in variants with decreasing L or if L<<50 just breaks down.

[cjain7]

I did observe a general increase while setting values 50, 10 and 2.

[lh3]

I never thought to support single base changes. Probably some part of the code makes incorrect assumptions and fails given -L 1. Need to check.