fast-nnt: fast, reproducible, and scalable neighbour network analysis in R, Python, and CLI

Area of Science:

• Computational Biology
• Phylogenetics
• Bioinformatics

Background:

• Neighbour-net methods are crucial for visualizing complex evolutionary histories involving reticulation, admixture, or hybridization.
• Current implementations are often GUI-based, hindering reproducibility and integration into automated computational workflows.
• Existing tools face scalability issues, being slow and memory-intensive, thus limiting analyses to smaller datasets.

Purpose of the Study:

• To develop a computationally efficient and memory-lean reimplementation of neighbour-net algorithms.
• To provide interfaces for common programming languages (R, Python) and command-line usage.
• To overcome the computational bottlenecks of existing neighbour-net software for large-scale phylogenetic analyses.

Main Methods:

• Reimplementation of neighbour-net algorithms from SplitsTree4/6 in Rust.
• Development of interfaces for R (fastnntr) and Python (fastnntpy).
• Independent selection of circular ordering algorithms (Multi-Way, Closest-Pair) and split weight inference methods (Conjugate Gradient, Active-Set).

Main Results:

• fast-nnt demonstrates substantial speed and memory improvements over existing tools, analyzing 3,333 taxa 11x faster than SplitsTree6 with less than half the memory.
• The tool accepts any symmetric distance matrix and achieves near-identical accuracy to SplitsTree.
• Modular design allows for reproducible analyses by enabling independent selection of algorithms and methods.

Conclusions:

• fast-nnt removes significant computational barriers, enabling routine use of neighbour-net methods on large datasets.
• The software facilitates scalable and reproducible neighbour-network inference.
• This tool enhances the practicality of advanced phylogenetic visualization for modern, large-scale biological data.