Releases: xinehc/melon
Releases · xinehc/melon
v0.2.0
Added
- [breaking] Add decoy protein sequences (RefSeq fungi, protozoa, viral, plant, and human GRCh38/hg38) which effectively trap non-prokaryotic reads and prevent them from inflating total prokaryotic genome copy estimates if the pre-filtering module (default with
Kraken2
) is not enabled. Pre-filtering is no longer necessary even if samples are contaminated with human DNA or other common eukaryotes/viruses, unless the mean genome size of prokaryotes needs to be estimated. See 8918168 for more details. This function requires a database released on or after 2024-06-28.
Changed
- Simplify filtering criteria for alignments.
v0.1.6
v0.1.5
v0.1.4
v0.1.3
Changed
- Change alignment filtering criteria: make
AS
cutoff more stringent, dropMS
. See 7cc6dbd for details.
Fixed
- Fix a bug causing total genome copies not being properly calculated with diamond>=2.1.9.
Added
- Output both gap-compressed and gap-uncompressed (BLAST-like) identity.
- Refine output format.
v0.1.2
v0.1.1
Added
- Add options to control EM early stop.
Changed
- Use
scipy.sparse
to reduce peak memory usage and computational time. - Change default terminal condition of EM (
max_iteration
: 100 -> 1000;epsilon
: 1e-5 -> 1e-10).
Fixed
- Fix a bug causing chimeric reads not being aggregated.
v0.1.0
Added
- Output a json file to indicate the lineage of processed reads.
Changed
- Make databases indexed by default.
- Use only kraken's prediction for removal of non-prokaryotic reads.
- Change
--db_kraken
to--db-kraken
for consistency. - Change
sp[0-9]+
tosp[0-9]+
for consistency. - Change
copies
tocopy
in output files for consistency.
Fixed
- Prevent numpy from using all logical cores.