Skip to content

Miniprot-0.7 (r207)
Compare
Choose a tag to compare
@lh3 lh3 released this 26 Dec 03:21
· 60 commits to master since this release
v0.7
2e57d1a

Notable changes:

  • Improvement: replaced open syncmers with modimers. This simplifies the code
    and slightly reduces the memory at comparable k-mer sampling rate. This
    changes the index format.

  • Improvement: fine tune parameters for higher sensitivity at a minor cost of
    junction accuracy: a) only index ORFs >= 30bp; b) reduced max k-mer
    occurrences from 50k to 20k; c) sample k-mers at a rate of 50%; d) reduced
    min number of k-mers from 5 to 3; e) add a bonus chaining score for anchors
    on the same reference block.

  • Improvement: adjust the max k-mer occurrence dynamically per protein.

  • Improvement: implemented 2-level chaining like minimap2 and minigraph. This
    reduces chaining time.

  • Bugfix: fixed a rare off-by-1 memory violation

  • Bugfix: fixed a memory leak

Overall, miniprot becomes faster at slightly higher peak memory usage. It is
more sensitive to distant homologs, though the junction accuracy of additional
alignment is usually lower. Also importantly, the index format of miniprot has
been changed. Miniprot will throw an error if you use miniprot with pre-built
indices generated with older versions.

(0.7: 25 December 2022, r207)

Assets 4
Loading