Repository navigation
Releases
Releases · AADavin/ZOMBI2
v0.63.0
Added
- Nucleotide genome runs now accept a codon model as
model, including--model gy94 --omega 0.1on the command line. The codon model uses each coding gene's reading frame and strand, reading the minus strand as its reverse complement. ZOMBI2 checks DNA, not annotation: a coding gene spans whole codons, ends in a stop codon, and has no earlier stop codon. A gene that is not coding evolves underintergene_modelat the gene's speed. The codon model holds each coding gene's stop codon, each base shared by overlapping genes, and every codon touching a shared base.result.coding_genesreports coding status, andsequences_summary.jsonreportscodingandnot_codingcounts undergenes. (#501)
Fixed
- Fixed an
AttributeErrorin nucleotide genome runs on the parallel engine (parallel=,--parallel) whenintergene_modelwas omitted. These runs now use the default spacer model,jc69, as the serial engine always did. (#501)
v0.62.0
Added
- Added
replacement=and--replacementat nucleotide resolution for replacing transfers. A replacing transfer overwrites the recipient's homologue with the donor's stretch. A replacing transfer carries everything inside the donor's stretch and gives the recipient the donor's length. Withself_transfer=True, a transfer uses another homologue in the same genome or stays additive. The sequence level needs no option and evolves a replaced stretch from the donor's sequence at the moment of transfer. Runs withreplacement=0remain unchanged byte for byte. A transfer rate that falls with sequence divergence remains unbuilt, whiletransfer_to="distance"uses time. (#500)
v0.61.0
Added
- Added
gy94, Goldman and Yang's (1994) codon model, for gene families. Runs write DNA without stop codons.omegarequires an explicit value to set selection on the protein. The codon model supports genetic codes 11 and 4. A branch length measures substitutions per base.lengthcounts bases in multiples of 3. (#499)
v0.60.0
Added
simulate_sequencesnow acceptsfounding=(--founding) andfamily_multipliers=(--family-multipliers) for family and ordered genome runs. Both accept files written by an earlier run. Both reject nucleotide genome runs, which take starting DNA through genome-levelgff=andfasta=. Runs that pass neither argument remain unchanged. (#497)
v0.59.1
Fixed
- Fix gene tree counting in
--resolution nucleotideruns after a duplication or transfer of a block that a deletion or an insertion had already cut into pieces. The gene tree now has one leaf for each copy a genome holds. This also corrects block tree outputs, phylograms, and duplication and transfer counts ingenome_events.tsvandgenome_summary.json. The fix leaves genomes, alignments, and runs without indels unchanged. (#496)
v0.59.0
Changed
- GFF input now preserves overlapping genes by default, with genes that overlap sharing one block. Previously, overlaps raised an error unless trimming was enabled.
trim_overlaps=Truekeeps the old trimming behaviour. Usegene_span_in_block(family)to locate a gene within its block's alignment. A run without overlapping genes remains unchanged, byte for byte. (#495)
v0.58.1
Fixed
- With
--recphylo extantorextant_reconciliation, the projection now keeps the transfer with a loss when the donor's copy dies on a visible branch. Previously, the projection moved the transfer and skipped the loss on the donor branch and at speciations above it. Loss counts increase inrecphylo_fam<f>_true.xml,recphylo_fam<f>_recoverable.xml, and thelossescolumn offamily_origins.tsv. (#491)
v0.58.0
Changed
simulate_genomes_familywith a rate that differs between lineages, through a driver (scaled_by,set_by) or a per-lineage draw, no longer re-evaluates that rate on every lineage at every event. Each lineage's weight is kept and rebuilt when its own genome or its own driver changes, on the serial engine and on the per-family one (parallel=,stream_to=), origination included. A trait-driven run of 200 tips takes 0.8 s where it took 24 s, and its cost per event no longer grows with the tree. A run that reads a driver now adds those weights in pairs, so event times can differ in their last digit and the same seed can give a different run than it did; with the sums taken left to right it is byte for byte the 0.57.1 run. A run that reads no driver is unchanged. (#488)
v0.57.1
Fixed
- The lines the command line prints after a run (
→ run.zombi2 ← open this first) swap their arrows and dots for ASCII where the console cannot encode them, instead of ending a finished run with a'charmap' codecerror on a Windows console. (#486)
v0.57.0
Added
- The published profile mixtures C10 to C60 as a model decoration:
lg().profile_mixture("C60")(--mixture C60on the command line) gives every site of every family one profile's amino-acid frequencies, drawn by weight, over the model's exchangeabilities, the site-heterogeneousLG+C60of the field; it composes with.across_sites(gamma_shape=…)in either order (LG+C60+G4) and runs on the parallel engine and on streamed runs. The model carries the profiles asprofile_modelsandprofile_weights; the run records which profile each site drew inresult.profile_assignmentsandprofile_assignments.tsv(family,site,profile_number; tokenassignments). Refused withprofiles=,partitions=, a per-clade set, indels and a nucleotide run. A validation check pins the profile shares to the weights and each profile's sites to its frequencies. (#483)
Fixed
- A recorded run (
record=True) over several partitions logged each substitution's site counted from the start of its partition; it now counts from the start of the sequence. Recording withpartitions=stays refused, so no run that was accepted before is affected. (#483)