Repository files navigation

# compile and installgit clone https://github.com/lh3/unimapcd unimap&& make# simple use cases (-b24 below saves memory for this toy example)./unimap -b24 -c test/MT-human.fa test/MT-orang.fa> out.paf# prebuild index; recommended as unimap indexing is slower than minimap2 indexing./unimap -b24 -d MT-human.umi test/MT-human.fa./unimap -c MT-human.umi test/MT-orang.fa> out.paf# use presets (no test data)unimap -cxasm5 --cs -t8 ref.fa asm.fa# if asm.fa is near identical to ref.faunimap -cxhifi --cs -t8 ref.fa hifi-reads.fa# HiFi reads to reference alignmentunimap -cxont  --cs -t8 ref.fa ont-reads.fa# Nanopore reads to reference alignment

Unimap is a fork ofminimap2 optimized for assembly-to-referencealignment. It integrates theminigraph chaining algorithm and can alignthrough long INDELs (up to 100kb by default) much faster than minimap2. Unimapis a better fit for resolving segmental duplications and is recommended overminimap2 for alignment between high-quality assemblies.

Unimap does not replace minimap2 for other types of alignment. It drops thesupport of multi-part index and short-read mapping. Its long-read alignment isdifferent from minimap2 but is not necessarily better. Unimap is more of aspecialized minimap2 at the moment.

• With the defaultasm5 preset, unimap may align a highly diverged regionas a long insertions followed by a long deletion.Truvari may identifytwo false positive calls in this case, but these arguably are not errors.

• Unimap takes ~5 minutes to index a human genome, slower than minimap2. It isrecommended to save the index for faster startup.

• The defaultont preset has been tuned for more recent Nanopore reads at~95% accuracy.