ModelTeller: model selection for optimal phylogenetic reconstruction using machine learning

Abadi, Shiran; Avram, Oren; Pupko, Tal; Mayrose, Itay

doi:10.1101/2020.01.09.899906

Cited by 8 publications

(14 citation statements)

References 80 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Yet another alternative approach is to train a machine-learning algorithm to assess historical signal. Machine learning has recently been proposed in phylogenetics for substitution model selection (Abadi et al 2020), inference of tree topology (Suvorov et al 2019), species delimitation (Derkarabetian et al 2019), and analyses of molecular rates across lineages (Tao et al 2019). A random forest or an artificial neural network might prove to be highly effective for identifying the factors that are associated with accurate inferences.…”

Section: Discussionmentioning

confidence: 99%

Excluding loci with substitution saturation improves inferences from phylogenomic data

Duchêne

Mather

Wal

et al. 2021

Preprint

View full text Add to dashboard Cite

The historical signal in nucleotide sequences becomes eroded over time by substitutions occurring repeatedly at the same sites. This phenomenon, known as substitution saturation, is recognized as one of the primary obstacles to deep-time phylogenetic inference using genome-scale data sets. We present a new test of substitution saturation and demonstrate its performance in simulated and empirical data. For some of the 36 empirical phylogenomic data sets that we examined, we detect substitution saturation in around 50% of loci. We found that saturation tends to be flagged as problematic in loci with highly discordant phylogenetic signals across sites. Within each data set, the loci with smaller numbers of informative sites are more likely to be flagged as containing problematic levels of saturation. The entropy saturation test proposed here is sensitive to high evolutionary rates relative to the evolutionary timeframe, while also being sensitive to several factors known to mislead phylogenetic inference, including short internal branches relative to external branches, short nucleotide sequences, and tree imbalance. Our study demonstrates that excluding loci with substitution saturation can be an effective means of mitigating the negative impact of multiple substitutions on phylogenetic inferences.

show abstract

Section: Discussionmentioning

confidence: 99%

Excluding loci with substitution saturation improves inferences from phylogenomic data

Duchêne

Mather

Wal

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…The only other tool currently employing a machine learning approach for model estimation is ModelTeller, which uses random forests for identifying the correct model of sequence evolution (Abadi et al 2020). However, ModelTeller and our approach are sufficiently different in aim and methodology to make direct comparison meaningless.…”

Section: Discussionmentioning

confidence: 99%

ModelRevelator: Fast phylogenetic model estimation via deep learning

Burgstaller-Muehlbacher

Crotty

Schmidt

et al. 2021

Preprint

View full text Add to dashboard Cite

Selecting the best model of sequence evolution for a multiple sequence alignment (MSA) constitutes the first step of phylogenetic tree reconstruction. Common approaches for inferring nucleotide models typically apply maximum likelihood (ML) methods, with discrimination between models determined by one of several information criteria. This requires tree reconstruction and optimisation which can be computationally expensive. We demonstrate that neural networks can be used to perform model selection, without the need to reconstruct trees, optimise parameters, or calculate likelihoods. We introduce ModelRevelator, a model selection tool underpinned by two deep neural networks. The first neural network, NNmodelfind, recommends one of six commonly used models of sequence evolution, ranging in complexity from JC to GTR. The second, NNalphafind, recommends whether or not a Γ--distributed rate heterogeneous model should be incorporated, and if so, provides an estimate of the shape parameter, ɑ. Users can simply input an MSA into ModelRevelator, and swiftly receive output recommending the evolutionary model, inclusive of the presence or absence of rate heterogeneity, and an estimate of ɑ. We show that ModelRevelator performs comparably with likelihood-based methods over a wide range of parameter settings, with significant potential savings in computational effort. Further, we show that this performance is not restricted to the alignments on which the networks were trained, but is maintained even on unseen empirical data. ModelRevelator will be made freely available in the forthcoming version of IQ-Tree (http://www.iqtree.org), and we expect it will provide a valuable alternative for phylogeneticists, especially where traditional methods of model selection are computationally prohibitive.

show abstract

“…While it may be possible to ameliorate the influence of MSA uncertainty on relative model selection, we must also ask: Do we need to mitigate this issue in the first place? For example, recent studies have shown that, for both nucleotide and amino-acid models, the model selection procedure itself may not be a critical step in phylogenetic reconstruction, since different models with extreme differences in relative fit may not actually result in systematically different results (Spielman and Kosakovsky Pond 2018;Abadi et al 2019;Spielman 2020) although how the precise model used may influence branch length and/or divergence estimation remains an important question Abadi et al (2019Abadi et al ( , 2020. As such, if distinct models may yield highly similar inferences, optimizing the model selection procedure itself has diminishing returns, akin to optimizing a mouse trap in a house without mice.…”

Section: Discussionmentioning

confidence: 99%

“…We note that one limitation of this study is that we only explore the influence of MSA uncertainty on relative model selection and not other approaches to identifying best-fitting models, including tests of model adequacy (Goldman 1993a,b;Duchêne et al 2018), Bayesian assessments of absolute model fit (Brown 2014;Lewis et al 2014), or more recently-developed machine-learning methods for model selection (Abadi et al 2020). Indeed, it may be possible that other approaches to model selection are more robust to MSA uncertainty, but the computational demands of these methods prohibit similar large-scale benchmarking.…”

Section: Discussionmentioning

confidence: 99%

Relative model selection of evolutionary substitution models can be sensitive to multiple sequence alignment uncertainty

Spielman

Miraglia

2021

Preprint

View full text Add to dashboard Cite

Multiple sequence alignments (MSAs) represent the fundamental unit of data inputted to most comparative sequence analyses. In phylogenetic analyses in particular, errors in MSA construction have the potential to induce further errors in downstream analyses such as phylogenetic reconstruction itself, ancestral state reconstruction, and divergence estimation. In addition to providing phylogenetic methods with an MSA to analyze, researchers must also specify a suitable evolutionary model for the given analysis. Most commonly, researchers apply relative model selection to select a model from candidate set and then provide both the MSA and the selected model as input to subsequent analyses. While the influence of MSA errors has been explored for most stages of phylogenetics pipelines, the potential effects of MSA uncertainty on the relative model selection procedure itself have not been explored. In this study, we assessed the consistency of relative model selection when presented with multiple perturbed versions of a given MSA. We find that while relative model selection is mostly robust to MSA uncertainty, in a substantial proportion of circumstances, relative model selection identifies distinct best-fitting models from different MSAs created from the same set of sequences. We find that this issue is more pervasive for nucleotide data compared to amino-acid data. However, we also find that it is challenging to predict whether relative model selection will be robust or sensitive to uncertainty in a given MSA. We find that that MSA uncertainty can affect virtually all steps of phylogenetic analysis pipelines to a greater extent than has previously been recognized, including relative model selection.

show abstract

ModelTeller: model selection for optimal phylogenetic reconstruction using machine learning

Cited by 8 publications

References 80 publications

Excluding loci with substitution saturation improves inferences from phylogenomic data

Excluding loci with substitution saturation improves inferences from phylogenomic data

ModelRevelator: Fast phylogenetic model estimation via deep learning

Relative model selection of evolutionary substitution models can be sensitive to multiple sequence alignment uncertainty

Contact Info

Product

Resources

About