RAxML-NG v2.0-beta3

New Features:

• new --fast mode: simplified tree search heuristic with early stopping -> details
• automatic model selection: FreeRate support, better heuristics, efficient parallelization, checkpointing -> details
• model optimization: optionally use expectation maximization (EM) instead of L-BFGS-B to optimize FreeRate parameters (faster, but less accurate) -> deatils
• VCF input support (optional)
• new support metric: Gene concordance factor (gCF) -> details

Optimizations and Improvements:

• allow mulltifurcated trees with -rfdist and -support commands
• optionally use parsimony-based SPR rounds to improve parsimony score after stepwise addition
• option to reset branch lengths in user starting trees: --extra brlen-reset
• support phased genotypes in PHYLIP/FASTA
• new protein models: FLAVI, mtMET, mtINV, mtVER
• new genotype models: GT16 for phased genotypes

Bugfixes:

• pythia: fix feature computation for partitioned MSAs (minor effect on predicted difficulty)
• fix parallelization auto-config with multiple user starting trees
• FASTA: remove leading spaces in taxon names
• genotype model (GT10/GT16): fix double-mutation rate to the minimum value
• parallel parsimony: fix inflated memory consumption on large datasets

Full Changelog: 2.0-beta2...2.0-beta3

RAxML-NG v2.0-beta2

Main new features

• Adaptive tree search heuristic (Togkousidis 2023) , (Togkousidis 2024)
• Multiple fast branch support measures
• Integrated model testing

Full Changelog: 1.2.2...2.0-beta2

RAxML-NG v1.2.2

Bugfixes

• --evaluate: use lh_epsilon=0.1 by default
• fix yet another bug with incomplete constraint trees
• improve CPU core count detection in node-sharing scenario

Other

Starting with v1.2.2, we provide a universal macOS binary which should work on both old Intel/x86 and new Apple/Arm chips.
We strongly recommend using this pre-compiled binary on macOS, since alternative installation methods (homebrew, compilation from source) might lead to hard-to-debug errors!

RAxML-NG v1.2.1

Bugfixes

• please see commit history

RAxML-NG v1.2.0

This release introduces multiple performance optimizations which yield 2x to 5x speedup for most typical analyses and datasets.

IMPORTANT: Due to change in defaults, the results are not compatible with previous raxml-ng versions. If you need to reproduce the behavior of raxml-ng 1.1.0 , add --extra compat-v11. Most changes can be also disabled one-by-one, please see instructions below:

New defaults & optimizations

• --search1: use parsimony starting tree by default
• bootstrap: use parsimony starting trees by default (add --extra bs-start-rand to use random starting trees as before)
• faster CLV updates in SPR rounds by @togkousa (#157), add --extra fastclv-off to disable
• new logLH epsilon defaults by @tschuelia: eps=1000 for brlen triplet optimization, eps=0.1 for final model optimization, eps=10 everywhere else (details: https://www.biorxiv.org/content/10.1101/2022.07.13.499893v1.full.pdf)
• parallelize parsimony starting tree computation (#151)

New features

• new amino-acid replacement matrices: Q.pfam, Q.bird, Q.insect, Q.mammal, Q.plant, Q.yeast (details: https://doi.org/10.1093/sysbio/syab010)
• arm64 support incl. vectorization via sse2neon library (#142)

Bugfixes

• fix regression failure with CATG input (#131)
• do not auto-enable rate scalers with >2000 taxa if --force / --force model_rate_scalers is specified (#133)
• partially incorrect TBE support values in --all mode
• --consense: fix error if result is a star tree (#143)
• invalid buffer size which led to failures in some parallel configurations (MPI with multiple workers per rank)
• truncate FASTA sequence names after the first space by default (#90). Alternatively, spaces can be converted to underscores with --msa-format fasta_longlabels.

Packaging note

This release is built with pll-modules 0.4.0, and libpll-2 0.4.0, and terraphast-0.1.0.

RAxML-NG v1.1.0

This release introduces several features/bugfixes that improve speed and/or accuracy of constrained tree search, especially in case of an incomplete constraint tree.

Constrained tree search improvements

• switch to a new constraint check approach based on split compatibility (old method still available via --extra constraint-old)
• allow user and parsimony-based starting trees in combination with topological constraint
• use 10 random + 10 parsimony starting trees by default (previously: 20 random)
• --check/--parse will check compatibility with constraint if specified (eg raxml-ng --check --msa aln.fa --model GTR+G --tree-constraint cons.nw --tree tree2check.nw)

Other

• add --bs-write-msa flag to write out bootstrap MSAs from main inference run (PR #125, thanks to @harmsm)
• fix assertion when rooting a multifurcating tree (#130)

RAxML-NG v1.0.3

BUGFIXES:

• --consense: fix empty output file prefix set by default
• --sitelh: fix slow printing of per-site log-likelihoods
• fix issues when compiled without -D_RAXML_PTHREADS
• coarse-grained: check that number of threads is multiple of the number of workers (#120)
• fix memory corruption in FreeRate optimization
• fix numerical issues due to zero rates in mtMAM matrix
• fix invariant site detection to account for ambiguous states (#121)

RAxML-NG v1.0.2

BUGFIXES:

• check for square brackets in taxon names (not allowed by Newick)
• fix 'LIBPLL-111: memory exhausted' error on very large ladder-like trees
• fix parallel compilation of tests (#108)
• do not print incorrect energy measurements after the latest Linux kernel patch
• coarse-grained parallelization: fix bootstrap failure/hanging in some special cases (eg autoMRE{x} where x<50)
• check for unifurcations in constraint tree (#114)
• improve error message when invalid number of substitution rates is specified

RAxML-NG v1.0.1

BUGFIXES:

• regression: RF computation with multiple files (--rf tree1,tree2) reported incorrect, unrealistically high distances (eg >0.8). However, --rf treeset.nw format was not affected and reported correct values.
• load balancing failed in rare edge cases (<=1 sites/core)

RAxML-NG v1.0.0

Major new features

• coarse-grained parallelization across tree searches (--workers option) -> see details
• adaptive parallelization (--threads auto{MAX}, --workers auto) -> see detais
• print per-site log-likelihoods (--sitelh command, #26)
• alignment column weighting (--site-weights option, #51) -> see details

Other features & optimizations

• energy monitoring via RAPL -> see details
• partial loading of RBA alignments (ExaML-style, saves memory for MPI runs on large supermatrices)
• detect CPU core oversubscription to prevent extreme slowdowns -> see details
• collapse short branches (bl=min) into multifurcations (#23)
• write down intermediate results (mlTrees, bootstraps)
• much faster (20x-30x) parsimony starting tree generation

Bugfixes

• asc bias failed under certain circumstances (site repeats disabled and non-DNA data)
• fix zero subst. rates in FLU model (#93)
• out-of-range branch lengths after scaling/normalization (eg #86)
• base frequency optimization sporadically failed when estimates hit the boundaries

Minor

• print CPU model/RAM/core count in logfile
• if --blmin is specified, adjust --precision accordingly
• remove BS support + 0.0 brlen at the root
• ensure that min/max brlens are not exceeded in user starting tree
• check that --msa file is not empty and not a directory
• check if start trees are bifurcating (#96)
• check if partition range has all sites in single-partition case
• check if +BU{} is specified in single-partition case
• do not use word "pattern" if pattern compression is off