Unblended disjoint tree merging using GTM improves species tree estimation

Abstract

Background: Phylogeny estimation is an important part of much biological research, but large-scale tree estimation is infeasible using standard methods due to computational issues. Recently, an approach to large-scale phylogeny has been proposed that divides a set of species into disjoint subsets, computes trees on the subsets, and then merges the trees together using a computed matrix of pairwise distances between the species. The novel component of these approaches is the last step: Disjoint Tree Merger (DTM) methods. Results: We present GTM (Guide Tree Merger), a polynomial time DTM method that adds edges to connect the subset trees, so as to provably minimize the topological distance to a computed guide tree. Thus, GTM performs unblended mergers, unlike the previous DTM methods. Yet, despite the potential limitation, our study shows that GTM has excellent accuracy, generally matching or improving on two previous DTMs, and is much faster than both. Conclusions: The proposed GTM approach to the DTM problem is a useful new tool for large-scale phylogenomic analysis, and shows the surprising potential for unblended DTM methods.The estimation of evolutionary trees (i.e., phylogenies, whether of genes or of species) is a fundamental step in much biological research, including understanding how species adapt to their environments, how gene function evolves, and how humans migrated across the globe. Yet, phylogeny estimation is computationally intensive, as nearly all the best approaches are based on NP-hard optimization problems.Divide-and-conquer approaches to tree estimation have the potential to provide improved scalability to heuristics for NP-hard optimization problems, but they depend on supertree methods, which have not been established to be highly scalable [1]. A new divide-and-conquer approach

Phylogeny Estimation Given Sequence Length Heterogeneity

“…While estimating phylogenies from unaligned datasets is very difficult for ultra-large datasets, there has been substantial progress over the last few years that suggests that dataset size is not likely to remain a significant impediment in the long term. For example, there are divide-and-conquer strategies for improving scalability of phylogeny estimation methods (e.g., TreeMerge ( Molloy and Warnow, 2019 ) and Guide Tree Merger ( Smirnov and Warnow, 2020 )) that do not require aligned sequence inputs and that could be used with any tree estimation method, including computationally intensive methods (e.g., Bayesian MCMC) or maximum likelihood estimation under complex models (e.g., the GHOST model ( Crotty et al, 2020 ) available in IQtree). These and future advances may make it feasible to estimate highly accurate trees from ultra-large datasets of unaligned sequences without burdensome computational requirements.…”

BATCH-SCAMPP: Scaling phylogenetic placement methods to place many sequences

“…Studies have shown that scalable phylogeny estimation methods have suffered in the presence of sequence length heterogeneity, and phylogenetic placement may provide a more accurate alternative [13]. For example, the divide-and-conquer tree estimation pipeline GTM [24] could benefit by using BSCAMPP. BSCAMPP can facilitate the initial tree decomposition of GTM for better placement of shorter, fragmentary sequences into an initial tree containing the longer full-length sequences, potentially leading to better final tree estimation.…”

Performance Evaluation and Parametric Optimization of Coal-Fired Water Tube Boiler Using the Grey-Taguchi Method

“…The gray partnership generation sequences are shown in Tables 3-5. The optimal sequence for SP, ST, and SSFR is x0 k k = 1, 2, ⋯ ⋯ , 7 [23]. The grey relational grade description in the grey relational analysis demonstrates how the seven sequences (x0 k and xi k , if = 1, 2) are related to the grade, 7, k = 1, 2, 7 [24].…”