CoChat
Sign Up
Cited by 780Genomics and Phylogenetic StudiesRNA and protein synthesis mechanismsEvolution and Paleontology Studies

The DOI matches its Crossref record; 48 of the 49 references it cites were found in a registry or an index, and 1 are not checked yet.

Selecting optimal partitioning schemes for phylogenomic datasets

Open access
Lanfear, Robert, Calcott, Brett, Kainer, David, Mayer, Christoph, Stamatakis, Alexandros·2014·BMC Evolutionary Biology
DOI 10.1186/1471-2148-14-82PMC PMC4012149
Open this paper in CoChat — free accountOpen PDF

Abstract

BACKGROUND: Partitioning involves estimating independent models of molecular evolution for different subsets of sites in a sequence alignment, and has been shown to improve phylogenetic inference. Current methods for estimating best-fit partitioning schemes, however, are only computationally feasible with datasets of fewer than 100 loci. This is a problem because datasets with thousands of loci are increasingly common in phylogenetics. METHODS: We develop two novel methods for estimating best-fit partitioning schemes on large phylogenomic datasets: strict and relaxed hierarchical clustering. These methods use information from the underlying data to cluster together similar subsets of sites in an alignment, and build on clustering approaches that have been proposed elsewhere. RESULTS: We compare the performance of our methods to each other, and to existing methods for selecting partitioning schemes. We demonstrate that while strict hierarchical clustering has the best computational efficiency on very large datasets, relaxed hierarchical clustering provides scalable efficiency and returns dramatically better partitioning schemes as assessed by common criteria such as AICc and BIC scores. CONCLUSIONS: These two methods provide the best current approaches to inferring partitioning schemes for very large datasets. We provide free open-source implementations of the methods in the PartitionFinder software. We hope that the use of these methods will help to improve the inferences made from large phylogenomic datasets.

Overview

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1186/1471-2148-14-82

Cites

48 of 49 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 1 The strict clustering algorithm performs poorly, but the relaxed clustering algorithm performs almost as well as the greedy algorithm. All analyses were conducted on a phylogenomic dataset of birds (Table 1, Hackett_2008). Note that lower scores indicate a better fit of the model to the data. The dashed line in each plot shows the score of the best partitioning scheme found by the greedy algorithm. Each boxplot represent the distribution of scores for 1000 runs of the strict or relaxed clustering algorithms, where each run uses a different definition of the similarity of two subsets (see main text). The figure shows that the relaxed clustering algorithm’s performance approaches that of the greedy algorithm as P increases, and that analysing 10% of partitioning schemes results in information theoretic scores that are very close to that of the greedy algorithm.
Figure 1 The strict clustering algorithm performs poorly, but the relaxed clustering algorithm performs almost as well as the greedy algorithm. All analyses were conducted on a phylogenomic dataset of birds (Table 1, Hackett_2008). Note that lower scores indicate a better fit of the model to the data. The dashed line in each plot shows the score of the best partitioning scheme found by the greedy algorithm. Each boxplot represent the distribution of scores for 1000 runs of the strict or relaxed clustering algorithms, where each run uses a different definition of the similarity of two subsets (see main text). The figure shows that the relaxed clustering algorithm’s performance approaches that of the greedy algorithm as P increases, and that analysing 10% of partitioning schemes results in information theoretic scores that are very close to that of the greedy algorithm.CC BY · PubMed Central
Figure 2 The performance of the strict clustering algorithm varies dramatically depending on the weighting schemes used to define subset similarity. The Y axis shows the difference in the AICc or BIC score compared to the best scheme found by the strict hierarchical clustering algorithm on the phylogenomic dataset from birds (Table 1). The X axes show the weights assigned to each of four parameter classes used to define subset similarity. Each panel shows 1000 data points, where each datapoint represents a single run of the strict hierarchical clustering algorithm under a particular weighting scheme. The set of four weights under which the best scheme was found by the strict hierarchical clustering algorithm are shown in red.
Figure 2 The performance of the strict clustering algorithm varies dramatically depending on the weighting schemes used to define subset similarity. The Y axis shows the difference in the AICc or BIC score compared to the best scheme found by the strict hierarchical clustering algorithm on the phylogenomic dataset from birds (Table 1). The X axes show the weights assigned to each of four parameter classes used to define subset similarity. Each panel shows 1000 data points, where each datapoint represents a single run of the strict hierarchical clustering algorithm under a particular weighting scheme. The set of four weights under which the best scheme was found by the strict hierarchical clustering algorithm are shown in red.CC BY · PubMed Central
Figure 3 The relaxed clustering algorithm outperforms the strict clustering algorithm across the 10 datasets shown in Table1. All scores are standardised by the score increase achieved by the greedy algorithm (i.e. the score of the best partitioning scheme from the greedy algorithm minus the score of the starting scheme), so that performance can be compared across datasets. Thus, the greedy algorithm always scores 100%, and is shown only for reference. Each line connects the results from a single dataset, demonstrating that in all cases using both the AICc and the BIC, the greedy algorithm performed best, the relaxed clustering algorithm (with 10% of schemes analysed) performed second best, and the strict clustering algorithm performed the worst. All analyses use the RAxML version of PartitionFinder.
Figure 3 The relaxed clustering algorithm outperforms the strict clustering algorithm across the 10 datasets shown in Table1. All scores are standardised by the score increase achieved by the greedy algorithm (i.e. the score of the best partitioning scheme from the greedy algorithm minus the score of the starting scheme), so that performance can be compared across datasets. Thus, the greedy algorithm always scores 100%, and is shown only for reference. Each line connects the results from a single dataset, demonstrating that in all cases using both the AICc and the BIC, the greedy algorithm performed best, the relaxed clustering algorithm (with 10% of schemes analysed) performed second best, and the strict clustering algorithm performed the worst. All analyses use the RAxML version of PartitionFinder.CC BY · PubMed Central
Figure 4 The strict and relaxed clustering algorithms are computationally much more efficient than the greedy algorithm. This figure shows the time taken by the relaxed and strict hierarchical clustering algorithms on the 10 datasets shown in Table 1, relative to the time taken by the greedy algorithm. All times are standardised by the time taken by the greedy algorithm, so that performance can be compared across datasets. Thus, the greedy algorithm always scores 100%, and is shown only for reference. Each line connects the results from a single dataset. The results show that the relaxed clustering algorithm (with 10% of schemes analysed) consistently takes about 10% of the time taken by the greedy algorithm, and that the strict hierarchical clustering algorithm takes between around 1% to 20% of the time taken by the greedy algorithm, depending on the dataset. All analyses use the RAxML version of PartitionFinder.
Figure 4 The strict and relaxed clustering algorithms are computationally much more efficient than the greedy algorithm. This figure shows the time taken by the relaxed and strict hierarchical clustering algorithms on the 10 datasets shown in Table 1, relative to the time taken by the greedy algorithm. All times are standardised by the time taken by the greedy algorithm, so that performance can be compared across datasets. Thus, the greedy algorithm always scores 100%, and is shown only for reference. Each line connects the results from a single dataset. The results show that the relaxed clustering algorithm (with 10% of schemes analysed) consistently takes about 10% of the time taken by the greedy algorithm, and that the strict hierarchical clustering algorithm takes between around 1% to 20% of the time taken by the greedy algorithm, depending on the dataset. All analyses use the RAxML version of PartitionFinder.CC BY · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 48 references

Cited by

Sorted by citations

Related

See all

Metadata

Authors
Lanfear, R., Calcott, B., Kainer, D., Mayer, C., & Stamatakis, A.
Affiliations
  • Australian National University (AU)
  • Zoological Research Museum Alexander Koenig (DE)
  • Karlsruhe Institute of Technology (DE)
Corresponding author
Lanfear, Robert
ORCID
Venue
BMC Evolutionary Biology
Publisher
Springer Science and Business Media LLC
Published
2014
Volume
14
Issue
1
Pages
82
ISSN
1471-2148 (ISSN-L)
DOI
10.1186/1471-2148-14-82
PMC
PMC4012149
OpenAlex
W2029226703
Cited by
762
References
48
Citations by year
2026: 40 · 2025: 50 · 2024: 63 · 2023: 52 · 2022: 57 · 2021: 98 · 2020: 79 · 2019: 71 · 2018: 71 · 2017: 76 · 2016: 58 · 2015: 41 · 2014: 6
FWCI
24.75
Type
journal article
Language
en
Open access
Licence
cc-by
Locations
  • BMC Evolutionary Biology (publishedVersion, cc-by)
  • Europe PMC (PubMed Central) (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • Informationszentrum Lebenswissenschaften (ZB MED) (submittedVersion, cc-by)
  • ANU Open Research (Australian National University) (submittedVersion)
Funders
National Evolutionary Synthesis Center
Topics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Genomics and Phylogenetic Studies
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › RNA and protein synthesis mechanisms
  • Physical Sciences › Earth and Planetary Sciences › Paleontology › Evolution and Paleontology Studies
Keywords
partitioning schemes, phylogenomics, hierarchical clustering, PartitionFinder, AICc, BIC
MeSH terms
Algorithms, Phylogeny, Software, Cluster Analysis, Evolution, Molecular, Genomics

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review