Phylogenetic trees and Euclidean embeddings

TL;DR

This paper explains why a square root transformation of phylogenetic distances allows taxa to be embedded in Euclidean space, offering new insights into tree structures and algorithms like NJ and BIONJ.

Contribution

It provides a direct, elementary explanation for the Euclidean embedding of phylogenetic trees, enhancing understanding of tree structures and related algorithms.

Findings

Square root transformation enables Euclidean embedding of taxa.

The embedding offers new insights into tree-building algorithms.

Reinterprets differences between NJ and BIONJ methods.

Abstract

It was recently observed by de Vienne et al. that a simple square root transformation of distances between taxa on a phylogenetic tree allowed for an embedding of the taxa into Euclidean space. While the justification for this was based on a diffusion model of continuous character evolution along the tree, here we give a direct and elementary explanation for it that provides substantial additional insight. We use this embedding to reinterpret the differences between the NJ and BIONJ tree building algorithms, providing one illustration of how this embedding reflects tree structures in data.