Skip to content

v0.5.1 (4-5-2026) - Extra checkpoints, faster graph cleaning + bugfix

Major changes

Results may change slightly.

v0.5.0 (3-12-2026) - RAM usage + read-mapping optimizations + stalling bug

Major changes

  • Using huffman coding to reduce RAM further. Should optimize by ~10-20%.
  • Read-to-read mapping is accelerated by 3-4x in special cases. This should reduce stalling behavior on certain datasets with high-depth, recombinant populations. Results will change slightly.
  • Fixed a major bug where myloasm would stall during the polishing step. This was caused by base quality runs of PHRED score 33.

Minor changes

  • Logging improved slightly
  • (ALPHA) There's a way to optimize memory for the k-mer counting step slightly using kmc as a pre-assembly step. Will improve this; message me if your assembly is dying early.

v0.4.0 (1-28-2026) - At least 30% less peak RAM than before at baseline, but possibly up to ~80% reduced (i.e., even 5-fold) for large metagenomes

Major changes

  • Data structure changes (liberal usage of variable-length quantity encodings) reduced memory by ~30-40% at baseline.
  • Reworking of read mapping logic: temporary mappings do not have to be stored in memory temporarily. For large metagenomes, this could reduce peak RAM by multiple folds.

Minor changes

  • Reworked the logic for removing contained reads slightly; it now proceeds in batches. This gave slightly better results on preliminary data.

BREAKING:

  • The checkpoints from previous runs for v0.3.0 and below will no longer work. That is, if you want to resume a disrupted v0.3.0 run using myloasm exist ..., you will have to use v0.3.0.

v0.3.0 (12-19-2025) - Better polishing for multi-strain communities & runtime on complex metagenomes

Major changes

  • Polishing may be much improved when multiple co-existing strains are assembled for nanopore data. Otherwise, results should not change much.
  • Bloom filter size is now automatically chosen by default (but can be overridden)
  • Significantly improved runtime on complex metagenomes (graph cleaning + read mapping steps)
  • Procedure for dereplicating spurious contigs prior to polishing changed slightly; should improve memory and time for complex metagenomes

Minor changes:

  • Added a new filter that removes poorly assembled, low-coverage contigs with mapping issues.
  • Added more options for dereplication of alternate contigs

v0.2.0 - More conservative and cleaner contig outputs with better polishing

  • allow fasta reads (https://github.com/bluenote-1577/myloasm/issues/3)
  • allow reads with non ACGT; send non-ACGT --> A character (https://github.com/bluenote-1577/myloasm/issues/4)
  • fixed some bugs with polishing. Should be more robust and have less ~200 bp gaps.
  • remove ALL contigs with cov <= 1 (cli parameter now)
  • remove singleton contigs with cov <= 3 (cli parameter now)
  • fixed specific polishing issues for small circular genomes
  • more aggressive dereplication of multiplied plasmids.
  • changed contig output format to output duplicated-yes|no|possibly. k-mer multiplicity is now in an extra field due to decimal being unruly especially during contig processing.