veryfasttree を Homebrew, apt, MacPorts, Nix でインストール

プロジェクトリンク

• ホームページ
• パッケージマネージャページ

コマンドとエイリアス

• VeryFastTree

プロジェクトの歴史

The project began publicly in 2019 and was introduced in a 2020 Bioinformatics paper by researchers at CiTIUS, Universidade de Santiago de Compostela. The paper framed VeryFastTree as a response to FastTree-2's limited scalability: FastTree-2 was already a successful large-phylogeny tool, but important maximum-likelihood rearrangement and posterior-distribution steps did not scale well across threads.

VeryFastTree v3.0 refactored and optimized the FastTree-2 code in C++ with OpenMP, added more complete vector-instruction support, parallelized major tree-improvement phases, and made same-thread-count parallel runs deterministic. The authors reported that a 330,000-sequence alignment could be processed in 4.5 hours on a standard server, substantially faster than FastTree-2 in their benchmark.

Version 4.0, described in a 2024 GigaScience paper, pushed the same idea to million-taxon datasets. It parallelized more tree traversal operations, including subtree pruning and regrafting moves, added compressed and new file-format support, improved compatibility, and introduced disk-computing functionality for users without enough RAM for the largest jobs.

採用の歴史

The project deliberately preserved FastTree-2's command-line arguments so researchers could replace a FastTree-2 invocation with `VeryFastTree` and keep the same workflow. That compatibility is an important adoption mechanism in bioinformatics, where command lines are often embedded in scripts, notebooks, and pipeline definitions.

The 2024 paper and project README point to research-community adoption by emphasizing availability, packaging in scientific software channels, and Python bindings. Its relevance grew with the scale of modern sequencing datasets, where tree-building for hundreds of thousands or millions of taxa is a practical bottleneck rather than an edge case.

使われ方

VeryFastTree is used to infer approximate maximum-likelihood phylogenies from nucleotide or protein alignments in formats such as FASTA, FASTQ, NEXUS, and PHYLIP. Users typically pass the same options they would pass to FastTree-2, adding VeryFastTree-specific controls when tuning thread count, deterministic mode, vector extensions, disk computing, or GPU-related experimental paths.

Its package-nerd appeal is that it is a drop-in performance replacement rather than a new workflow to learn. In a field full of heavyweight scientific tools, keeping the old command surface while changing the performance envelope is a practical way to get into existing pipelines.

パッケージ好きにとっての重要性

VeryFastTree matters because it turns a known scientific workhorse into something more suitable for current dataset sizes. It preserves the pragmatic FastTree-2 tradeoff of fast approximate maximum-likelihood inference, but makes that tradeoff viable on larger multicore servers and, with v4.0, on million-taxon alignments.

タイムライン

• 2019-12-02: GitHub repository created.
• 2020-06-23: Bioinformatics paper published introducing VeryFastTree for large alignments.
• 2020-06-25: v3.0 GitHub release published.
• 2023-06: v4.0 release introduced new thread levels and broader parallelization.
• 2024-08-08: GigaScience paper published describing VeryFastTree 4.0 and million-taxon benchmarks.

Related projects

• FastTree-2: the direct methodological and command-line ancestor.
• RAxML, PhyML, and IQ-TREE: other maximum-likelihood phylogeny tools discussed in the Bioinformatics paper as part of the large-tree inference landscape.
• VeryFastTree Python bindings: a related package for embedding the tool in Python workflows.

ソース

• https://academic.oup.com/bioinformatics/article/36/17/4658/5861530
• https://academic.oup.com/gigascience/article/doi/10.1093/gigascience/giae055/7730000
• https://api.github.com/repos/citiususc/veryfasttree
• https://api.github.com/repos/citiususc/veryfasttree/releases?per_page=100
• https://github.com/citiususc/veryfasttree