From Easy to Hopeless—Predicting the Difficulty of Phylogenetic Analyses

Haag, Julia; Höhler, Dimitri; Bettisworth, Ben; Stamatakis, Alexandros

doi:10.1093/molbev/msac254

Cited by 30 publications

(25 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For both GBT and CNN classifiers, we observed a general trend for lower classification accuracy on more difficult MSAs according to the Pythia difficulty score. The higher the Pythia difficulty for an MSA, the lower the signal in the data and the more difficult it is to obtain a well-supported phylogeny as the likelihood surface exhibits multiple indistinguishable (by means of standard phylogenetic significance tests) likelihood peaks [21]. In addition to assessing the BACC as a function of the difficulty of simulated MSAs, we also assessed the BACC as a function of the difficulty of the underlying empirical MSAs.…”

Section: Resultsmentioning

confidence: 99%

“…For data collections simulating indel events, we also used the proportion of gaps as feature ( % gaps ). Further, we quantified the signal in the MSA using the difficulty of the respective phylogenetic analysis as predicted by Pythia [21] ( difficulty ), as well as the Shannon entropy [44] of the MSA ( Entropy ), a multinomial test statistic of the MSA ( Bollback multinomial ; [8]), and an entropy-like metric based on the number and frequency of patterns in the MSA ( Pattern entropy ). For further details on the computation of these metrics, we refer the interested reader to Supplementary Material Section 4.1.…”

Section: Methodsmentioning

confidence: 99%

See 1 more Smart Citation

Simulations of sequence evolution: how (un)realistic they are and why

Trost

Haag

Höhler

et al. 2023

Preprint

Self Cite

View full text Add to dashboard Cite

Motivation: Simulating sequence evolution plays an important role in the development and evaluation of phylogenetic inference tools. Naturally, the simulated data needs to be as realistic as possible to be indicative of the performance of the developed tools on empirical data. Over the years, numerous phylogenetic sequence simulators, employing various models of evolution, have been published with the goal to simulate such empirical-like data. In this study, we simulated DNA and protein Multiple Sequence Alignments (MSAs) under increasingly complex models of evolution with and without insertion/deletion (indel) events using a state-of-the-art sequence simulator. We assessed their realism by quantifying how well supervised learning methods are able to predict whether a given MSA is simulated or empirical. Results: Our results show that we can distinguish between empirical and simulated MSAs with high accuracy using two distinct and independently developed classification approaches across all tested models of sequence evolution. Our findings suggest that the current state-of-the-art models fail to accurately replicate the process of evolution.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Methodsmentioning

confidence: 99%

Simulations of sequence evolution: how (un)realistic they are and why

Trost

Haag

Höhler

et al. 2023

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…We noticed that approximately half of the MSAs in TreeBASE contain at least two exactly identical sequences, and therefore decided to remove all duplicate sequences before training EBG. We selected the MSAs based on the Pythia difficulty [12]. Pythia quantifies the difficulty of phylogenetic analysis under the ML criterion.…”

Section: Training Datamentioning

confidence: 99%

“…RAxML-NG infers parsimony starting trees via a randomized stepwise addition order algorithm (−−start-option). The development of Pythia showed that by using the computationally substantially less expensive parsimony trees, we can accurately predict features of the ML tree space [12]. Due to the high prediction accuracy of Pythia, we therefore expect that parsimony-based features will also be useful for predicting SBS values.…”

Section: Feature Engineeringmentioning

confidence: 99%

Predicting Phylogenetic Bootstrap Values via Machine Learning

Wiegert,

Höhler,

Haag

et al. 2024

Preprint

Self Cite

View full text Add to dashboard Cite

Estimating the statistical robustness of the inferred tree(s) constitutes an integral part of most phylogenetic analyses. Commonly, one computes and assigns a branch support value to each inner branch of the inferred phylogeny. The most widely used method for calculating branch support on trees inferred under Maximum Likelihood (ML) is the Standard, non-parametric Felsenstein Bootstrap Support (SBS). Due to the high computational cost of the SBS, a plethora of methods has been developed to approximate it, for instance, via the Rapid Bootstrap (RB) algorithm. There have also been attempts to devise faster, alternative support measures, such as the SH-aLRT (Shimodaira-Hasegawa-like approximate Likelihood Ratio Test) or the UltraFast Bootstrap 2 (UFBoot2) method. Those faster alternatives exhibit some limitations, such as the need to assess model violations (UFBoot2) or meaningless low branch support intervals (SH-aLRT). Here, we present the Educated Bootstrap Guesser (EBG), a machine learning-based tool that predicts SBS branch support values for a given input phylogeny. EBG is on average 9.4 (σ = 5.5) times faster than UFBoot2. EBG-based SBS estimates exhibit a median absolute error of 5 when predicting SBS values between 0 and 100. Furthermore, EBG also provides uncertainty measures for all per-branch SBS predictions and thereby allows for a more rigorous and careful interpretation. EBG can predict SBS support values on a phylogeny comprising 1654 SARS-CoV2 genome sequences within 3 hours on a mid-class laptop. EBG is available under GNU GPL3.

show abstract

“…The main disadvantage is that an additional step is required in the analysis (either an MCMC search or the inference of a bootstrap distribution), and accurately computing these distributions can be challenging, especially on large numbers of taxa where the number of tips in the single gene trees can be large compared to the length of the multiple sequence alignment. The degree of difficulty for a phylogenetic inference on a given MSA or essentially lack of signal for obtaining a stable single gene (family) tree can now be predicted via machine learning methods (Haag et al 2022). However, because the approximation needs only to be computed once for each gene family, ALE is faster than GeneRax for the purpose of evaluating different rooted species trees (or root positions on a single unrooted topology).…”

Section: Inference Of Gene Duplication Transfer and Loss Events Using...mentioning

confidence: 99%

The power and limitations of species tree-aware phylogenetics

Williams

Davín

Morel

et al. 2023

Preprint

Self Cite

View full text Add to dashboard Cite

Species tree-aware phylogenetic methods model how gene trees are generated along the species tree by a series of evolutionary events, including the duplication, transfer and loss of genes. Over the past ten years these methods have emerged as a powerful tool for inferring and rooting gene and species trees, inferring ancestral gene repertoires, and studying the processes of gene and genome evolution. However, these methods are complex and can be more difficult to use than traditional phylogenetic approaches. Method development is rapid, and it can be difficult to decide between approaches and interpret results. Here, we review ALE and GeneRax, two popular packages for reconciling gene and species trees, explaining how they work, how results can be interpreted, and providing a tutorial for practical analysis. It was recently suggested that reconciliation-based estimates of duplication and transfer frequencies are unreliable. We evaluate this criticism and find that, provided parameters are estimated from the data rather than being fixed based on prior assumptions, reconciliation-based inferences are in good agreement with the literature, recovering variation in gene duplication and transfer frequencies across lineages consistent with the known biology of studied clades. For example, published datasets support the view that transfers greatly outnumber duplications in most prokaryotic lineages. We conclude by discussing some limitations of current models and prospects for future progress.

show abstract

From Easy to Hopeless—Predicting the Difficulty of Phylogenetic Analyses

Cited by 30 publications

References 29 publications

Simulations of sequence evolution: how (un)realistic they are and why

Simulations of sequence evolution: how (un)realistic they are and why

Predicting Phylogenetic Bootstrap Values via Machine Learning

The power and limitations of species tree-aware phylogenetics

Contact Info

Product

Resources

About