Evaluating the Performance of Probabilistic Algorithms for Phylogenetic Analysis of Big Morphological Datasets: A Simulation Study

Abstract

Reconstructing the tree of life is an essential task in evolutionary biology. It demands accurate phylogenetic inference for both extant and extinct organisms, the latter being almost entirely dependent on morphological data. While parsimony methods have traditionally dominated the field of morphological phylogenetics, a rapidly growing number of studies are now employing probabilistic methods (maximum likelihood and Bayesian inference). The present-day toolkit of probabilistic methods offers varied software with distinct algorithms and assumptions for reaching global optimality. However, benchmark performance assessments of different software packages for the analyses of morphological data, particularly in the era of big data, are still lacking. Here, we test the performance of four major probabilistic software under variable taxonomic sampling and missing data conditions: the Bayesian inference-based programs MrBayes and RevBayes, and the maximum likelihood-based IQ-TREE and RAxML. We evaluated software performance by calculating the distance between inferred and true trees using a variety of metrics, including Robinson-Foulds (RF), Matching Splits (MS), and Kuhner-Felsenstein (KF) distances. Our results show that increased taxonomic sampling improves accuracy, precision, and resolution of reconstructed topologies across all tested probabilistic software applications and all levels of missing data. Under the RF metric, Bayesian inference applications were the most consistent, accurate, and robust to variation in taxonomic sampling in all tested conditions, especially at high levels of missing data, with little difference in performance between the two tested programs. The MS metric favored more resolved topologies that were generally produced by IQ-TREE. Adding more taxa dramatically reduced performance disparities between programs. Importantly, our results suggest that the RF metric penalizes incorrectly resolved nodes (false positives) more severely than the MS metric, which instead tends to penalize polytomies. If false positives are to be avoided in systematics, Bayesian inference should be preferred over maximum likelihood for the analysis of morphological data.

Megaevolutionary dynamics and the timing of evolutionary innovation in reptiles

“…Three taxa that operate as wildcards in the present dataset, as identified in a previous study using the RogueNaRok algorithm 23,66 , were removed for the present analyses, namely Paliguana whitei, Palaeagama vielhaueri and Pamelina polonica. Increased taxon sampling and rogue taxon removal resulted in considerable improvement on resolution and convergency of results on early diapsid relationships between nonclock and clock trees (see Results)-as expected based on recent simulations data 67 .…”

Sphenodontian phylogeny and the impact of model choice in Bayesian morphological clock estimates of divergence times and evolutionary rates

“…Our tests were performed in Mr. Bayes given its greater stability when analyzed in high-performance computer clusters (RevBayes runs were often terminated prematurely in some tests). However, the extremely similar behavior of both software to analyze datasets varying in taxon number, amounts of missing data, and levels of character homoplasy [27] suggest that our conclusions here may extend to analyses with RevBayes as well.…”

“…The accumulation of taxa represented only by fragmentary remains increases the overall amount of missing information in the dataset, which reduces the ability of statistical methods to infer accurate phylogenies for morphological data, even at high levels of taxonomic sampling [27]. We thus focused on sampling species with more completely preserved specimens; our goal was to construct a robust general framework of sphenodontian relationships (into which more incomplete taxa may be added in the future by other researchers to address individual taxonomic problems).…”

“…This effectively implements in Mr. Bayes a diversity sampling strategy with fossils as tips only (not sampling for ancestors), and it is referred here as the "NoSA diversity" sampling strategy. We also note that a similar combination of model parameters can be performed in the sister program to Mr. Bayes, RevBayes [83], which has an overall similar performance for tree inference to Mr. Bayes using non-clock Bayesian inference for morphological datasets [27]. Our tests were performed in Mr. Bayes given its greater stability when analyzed in high-performance computer clusters (RevBayes runs were often terminated prematurely in some tests).…”

“…Recently, some studies have started to direct their attention to the performance of morphology only datasets in a probabilistic framework, especially so as to compare the performance of software and optimality criteria under various conditions using non-clock Bayesian inference-e.g., [23,[25][26][27][28]. However, very few studies have directed their attention to the performance of model choice using relaxed Bayesian clocks with morphological data only.…”