Reliable estimation of tree branch lengths using deep neural networks

Abstract

A phylogenetic tree represents hypothesized evolutionary history for a set of taxa. Besides the branching patterns (i.e., tree topology), phylogenies contain information about the evolutionary distances (i.e. branch lengths) between all taxa in the tree, which include extant taxa (external nodes) and their last common ancestors (internal nodes). During phylogenetic tree inference, the branch lengths are typically co-estimated along with other phylogenetic parameters during tree topology space exploration. There are well-known regions of the branch length parameter space where accurate estimation of phylogenetic trees is especially difficult. Several novel studies have recently demonstrated that machine learning approaches have the potential to help solve phylogenetic problems with greater accuracy and computational efficiency. In this study, as a proof of concept, we sought to explore the possibility of machine learning models to predict branch lengths. To that end, we designed several deep learning frameworks to estimate branch lengths on fixed tree topologies from multiple sequence alignments or its representations. Our results show that deep learning methods can exhibit superior performance in some difficult regions of branch length parameter space. For example, in contrast to maximum likelihood inference, which is typically used for estimating branch lengths, deep learning methods are more efficient and accurate when inferring long branches that are associated with distantly related taxa and perform well in the aforementioned challenging regions of the parameter space. Together, our findings represent a next step toward accurate, fast, and reliable phylogenetic inference with machine learning approaches.

Applications of Machine Learning in Phylogenetics

“…The dependence on the number of tips complicates the use of machine learning approaches. Suvorov and Schrider (2022) employed both a CNN and a multilayer perceptron (MLP) to infer branch lengths on fixed tree topologies with four or eight taxa. For the CNN-based approach, they adapted a previously proposed architecture (Suvorov et al, 2020).…”

Deep learning and likelihood approaches for viral phylogeography converge on the same answers whether the inference model is right or wrong

“…Once trained, neural networks have the benefit of being fast, easy to use, and scalable. Recently, likelihood-free deep learning neural network methods have successfully been applied to phylogenetics (Suvorov et al 2020;Suvorov and Schrider 2022b;Nesterenko et al 2022;Solis-Lemus et al 2022;da Fonseca et al 2020), and phylodynamic inference Lambert et al 2022).…”

“…;Schmitt et al (2022).With recent advances in deep learning in epidemiology, evolution, and ecology(Battey et al 2020;Schrider and Kern 2018;Radev et al 2021;Lambert et al 2022;Rosenzweig et al 2022;Suvorov and Schrider 2022a) biologists can now explore the behavior of entire classes of stochastic branching models that are biologically interesting but mathematically or statistically prohibitive for use with traditional likelihood-based inference techniques. Although we are cautiously optimistic about the future of deep learning methods for phylogenetics, it will become increasingly important for the field to diagnose the conditions where phylogenetic deep learning underperforms relative to likelihood-based approaches, and to devise general solutions for the field.Table S1: BEST comparisons between CNN and Bayesian absolute percent errors (APEs) for model parameters across all experiments.…”

AliSim-HPC: parallel sequence simulator for phylogenetics

“…However, many applications require simulations of a vast number of large MSAs. For example, training new machine learning applications for phylogenetics (Suvorov and Schrider, 2022; Burgstaller-Muehlbacher et al ., 2021; Abadi et al ., 2020; Suvorov et al ., 2020) requires millions of simulated MSAs. Due to its sequential implementation, AliSim becomes very slow, taking several days or weeks to simulate millions of alignments.…”