PaperlayerView exact revision on arXiv

Source-linked AI summary

DiscoVista: interpretable visualizations of gene tree discordance

Erfan Sayyari, James B. Whitfield, Siavash Mirarab

arXiv:1709.09305v2q-bio.PE

TL;DR

Phylogenomic analyses often disagree across genes and species-tree methods, but such conflicts can be difficult to interpret. DiscoVista generates simple visualizations of discordance and related dataset patterns; empirical examples reveal extensive gene-tree discordance and GC variation concentrated at third codon positions.

  • Problem

    Phylogenomic analyses show incongruence, yet interpretable ways to assess discordance among species trees and between gene trees and a summary species tree are needed.

    Source: §1 · pp. 1–2

  • Method

    DiscoVista generates visualizations of gene-tree and species-tree discordance, taxon occupancy, and sequence statistics, with options to focus analyses on groups of taxa.

    Source: §1 · p. 2

  • Results

    Empirical visualizations showed high gene-tree discordance despite mostly congruent ASTRAL and concatenation species trees, while plant GC variation was concentrated mostly at the third codon position.

    Source: §2 · p. 6

  • Takeaways & Limitations

    DiscoVista provides a tool for visualizing patterns of discordance, missing data, and GC variation in phylogenomic datasets.

    Source: §2 · p. 6

    Branch-quartet expectations assume the focal branch is correct, discordance arises only from ILS, and gene-tree estimation is error-free.

    Source: §2 · p. 6

Abstract

from arXiv · show

Phylogenomics has ushered in an age of discordance. Analyses often reveal abundant discordances among phylogenies of different parts of genomes, as well as incongruences between species trees obtained using different methods or data partitions. Researchers are often left trying to make sense of such incongruences. Interpretive ways of measuring and visualizing discordance are needed, both among alternative species trees and gene trees, especially for specific focal branches of a tree. Here, we introduce DiscoVista, a publicly available tool that creates a suite of simple but interpretable visualizations. DiscoVista helps quantify the amount of discordance and some of its potential causes.

1 Introduction

Phylogenomic studies face incongruence among gene trees and species-tree analyses, while noisy gene-tree estimates and methodological choices complicate interpretation. DiscoVista offers simple visualizations of discordance and related data patterns, with analyses focused on groups of taxa relevant to biological hypotheses.

1 Introduction

  • Discordance is common in phylogenomics, and noisy gene-tree estimates make its impacts difficult to assess.Methods addressing gene-tree incongruence exist, but no consensus has emerged on the best methodology.Source: §1 · p. 1
  • Researchers vary models, methods, species sets, and data partitions to test result sensitivity, but often observe incongruence among analyses.Ideally, all analyses would be published so readers can see that incongruence.Source: §1 · p. 1
  • Existing tree overlays and networks depict discordance, but their meaning can be difficult to interpret measurably.DiscoVista also summarizes gene-tree discordance, species-tree discordance, taxon occupancy, and GC content, with focal-group analyses to simplify interpretation.Source: §1 · pp. 1–2
  • DiscoVista creates simple visualizations to help systematists identify discordance and its potential causes.The command-line tool is designed to make these patterns interpretable and easy to assess.Source: §1 · p. 2

2 Results

Across the empirical datasets, DiscoVista visualizations reveal discordance among gene trees and species-tree analyses, alongside occupancy and sequence-composition patterns. Quartet frequencies around focal branches also show patterns relevant to evaluating discordance and its possible causes.

  • The examples apply DiscoVista to three datasets, including Xenacoelomorpha studies using different species sets and methods and a plant dataset with DNA and amino-acid sequences.The Xenacoelomorpha studies reached the same conclusion about the placement of Xenacoelomorpha as sister to all other Bilateria.Source: §2 · p. 2
  • Species-tree compatibility visualizations classify focal splits as supported, weakly rejected, or strongly rejected, with weak rejection assessed after contracting branches below a support threshold.Compatible splits can be added to the unrooted tree; two splits are compatible when at least one of their four pairwise intersections is empty.Source: §2 · p. 3
  • The two species-tree datasets are largely congruent, but aggressive occupancy filtering reduces concatenation-tree support and Chordata recovery in Rouse; ASTRAL on bestML trees seems more accurate than on MLBS trees.The compatibility figures compare focal splits across analyses.Source: §2 · p. 3, Figure 1 · p. 4
  • Occupancy plots report the fraction of alignments containing at least one species from a group, while occupancy heatmaps show species presence and relative non-gapped sequence length.The analyses use alignments and a species-annotation file.Source: §2 · p. 3
  • Occupancy patterns are fairly consistent across the Rouse dataset, whereas Cannon genes vary more; several species in both datasets have very low occupancy.In Rouse, low occupancy for Xenoturbella bocki may be ameliorated by high occupancy in Xenoturbella profoundus and Hofstenia miamia.Source: §2 · p. 3
  • Major splits are absent from most gene trees, although most gene trees are compatible with the species trees after contracting branches below 75% support.Rouse gene trees show somewhat less discordance with major species-tree splits than Cannon gene trees.Source: Figure 2 · p. 5, §2 · p. 6
  • For Nephrozoa and Deuterostomia, main quartet-topology frequencies are close to 1/3 in both studies, while alternative-topology frequencies fit MSC expectations better in Rouse than Cannon.The high discordance is likely caused by a combination of true discordance and gene-tree estimation error.Source: Figure 2 · p. 5, §2 · p. 6
  • In the plant dataset, GC variation is concentrated mostly at the third codon position, and the results favor removing that position in gene-tree or concatenation analyses.GC figures could not be computed for the Xenacoelomorpha datasets because they did not release DNA sequences.Source: §2 · p. 6

Supplementary Material for “DiscoVista: interpretable visualizations of gene

The supplement provides additional tables and figures for the DiscoVista paper. It identifies the authors and their institutional affiliations.

Supplementary Figures

The supplementary figures show DiscoVista outputs across the Xenoturbella and 1kp datasets. They cover split definitions, species- and gene-tree analyses, occupancy, quartet frequencies, and GC content.

  • Figures S1 and S2 give the complete split definitions for the Xenoturbella datasets of Rouse et al. (2016) and Cannon et al. (2016).Source: Figure S1 · p. 10, Figure S2 · p. 11
  • Figures S3 and S4 compare species-tree analyses across major orders and clades, methods, and two closely related 1kp datasets; S3 colors encode MLBS values, and weak rejection uses a 90% contraction threshold.Source: Figure S3 · p. 12, Figure S4 · p. 13
  • Figure S5 plots 1kp taxon occupancy by species and major clade under long-branch removal and fragmentary-sequence filtering conditions.Source: Figure S5 · p. 14
  • Figures S6 and S7 summarize 1kp gene-tree support, rejection, and missingness for important clades across three model conditions; weak rejection contracts branches below 75%.Source: Figure S6 · p. 15, Figure S7 · p. 16
  • Figure S8 shows frequencies of the three quartet topologies around focal ASTRAL species-tree branches, with dotted lines marking the 1/3 threshold.Source: Figure S8 · p. 17
  • Figure S9 presents 1kp GC-content boxplots and dot plots for individual codon positions and all three positions together.Source: Figure S9 · p. 18

Structure of parameter files

The supplement describes the split-definition and species-annotation files used to specify groups and species membership for DiscoVista analyses.

Installation, Commands, and Usage

The supplement gives installation guidance and Docker commands for DiscoVista analyses. The examples cover species-tree and gene-tree discordance, occupancy, GC content, and relative frequency analyses.

Loading 1709.09305v2…