QuartetNet: Novel Phylogenetic Quartet Tree Reconstruction Using Neural Networks

Abstract

Phylogenetic reconstruction is fundamental problem in bioinformatics that studies the origins and evolutionary of species. Traditionally, statistical inference methods have been used to reconstruct phylogentic tree from a Multiple Sequence Alignment (MSA) of species, visualising the evolutionary history using a tree structure. However, those methods are either inaccurate or time-consuming. Supervised machine learning, as an fast yet accurate alternative, can be used to reconstruct a phylogenetic tree by predicting the topology and branch lengths of the tree. In this study, we propose QuartetNet, a novel phylogentic reconstruction framework that uses neural networks (NNs) for reconstructing quartet (4-species) trees. Through comparative analysis, QuartetNet is shown to either match or surpass the performance of stat-of-the-art domain method. Show more