Abstract
Phylogenetic trees are essential for studying biology, but their reproducibility under identical parameter settings remains unexplored. Here, we find that 3515 (18.11%) IQ-TREE-inferred and 1813 (9.34%) RAxML-NG-inferred maximum likelihood (ML) gene trees are topologically irreproducible when executing two replicates (Run1 and Run2) for each of 19,414 gene alignments in 15 animal, plant, and fungal phylogenomic datasets. Notably, coalescent-based ASTRAL species phylogenies inferred from Run1 and Run2 sets of individual gene trees are topologically irreproducible for 9/15 phylogenomic datasets, whereas concatenation-based phylogenies inferred twice from the same supermatrix are reproducible. Our simulations further show that irreproducible phylogenies are more likely to be incorrect than reproducible phylogenies. These results suggest that a considerable fraction of single-gene ML trees may be irreproducible. Increasing reproducibility in ML inference will benefit from providing analyses’ log files, which contain typically reported parameters (e.g., program, substitution model, number of tree searches) but also typically unreported ones (e.g., random starting seed number, number of threads, processor type).
Highlights
Phylogenetic trees are essential for studying biology, but their reproducibility under identical parameter settings remains unexplored
To evaluate the reproducibility of single-gene phylogenetic trees, we collected 19,414 gene alignments from 15 animal, plant, and fungal phylogenomic data sets that span a wide spectrum of taxonomic ranks (Supplementary Data 1)
We found that collapsing branches with low bootstrap support values eliminated the irreproducibility of coalescentbased ASTRAL species phylogenies for three and four phylogenomic data sets when their gene trees were inferred by IQ-TEE and RAxML-NG, respectively, but eight phylogenomic data sets (IQ-TREE) and seven phylogenomic data sets (RAxML-NG) still yielded topologically different ASTRAL species phylogenies (Table 2)
Summary
Phylogenetic trees are essential for studying biology, but their reproducibility under identical parameter settings remains unexplored. A 2013 meta-analysis reported that phylogenetic trees in 6277/7539 (83.3%) studies published in the last few decades are irreproducible due to the unavailability of the underlying data[22]. The availability of public data repositories (e.g., TreeBASE, Dryad, Figshare, Zenodo, OSF) coupled with the modernization of journal data sharing policies have greatly increased the availability of sequence alignment data, the resulting phylogenetic trees, as well as of information about program(s) and key parameter settings (e.g., substitution model) used[24,25,26,27,28,29,30]. Phylogenetic studies seldom provide the random starting seed number used in inference, even though it is well established that it can affect hill-climbing tree heuristic searches of state-of-the-art, widely used maximum likelihood (ML)-based phylogenetic programs, such as IQ-TREE31 and RAxML-NG32. We currently know little about how the phylogenetic informativeness of the underlying data (e.g., the number of parsimony-informative sites or branch support values) or the variation in the computing resources used (e.g., number of the central processing unit (CPU) cores and type(s) of processor among studies or among nodes of a supercomputing cluster)[33,34] affect the reproducibility of phylogenetic inference
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
