PaperlayerView exact revision on arXiv

Source-linked AI summary

Agalma: an automated phylogenomics workflow

Casey W. Dunn, Mark Howison, Felipe Zapata

arXiv:1307.6432v1q-bio.GN

TL;DR

Phylogenomic analyses are complex and difficult to reproduce, compare, and extend. Agalma automates the workflow with detailed reports; its example analysis produced a preliminary species tree congruent with previous analyses of siphonophore relationships.

  • Problem

    Phylogenomic studies involve complex, computationally intensive steps, and manual or semiautomated analyses are difficult to reproduce, compare, and extend.

    Source: Background · p. 2

  • Method

    Agalma automates phylogenomic analyses from raw sequence reads to preliminary trees using high-level commands and detailed reports across data sets and analysis steps.

    Source: Background · p. 3

  • Results

    The preliminary species tree for the example analysis was congruent with previous analyses of siphonophore relationships.

    Source: Phylogenetic analyses · p. 8

  • Takeaways & Limitations

    Explicit analysis code enables reviewers and readers to reproduce the authors’ analysis without re-curating the dataset or rewriting the code.

    Source: Conclusions · p. 9

Abstract

from arXiv · show

In the past decade, transcriptome data have become an important component of many phylogenetic studies. Phylogenetic studies now regularly include genes from newly sequenced transcriptomes, as well as publicly available transcriptomes and genomes. Implementing such a phylogenomic study, however, is computationally intensive, requires the coordinated use of many complex software tools, and includes multiple steps for which no published tools exist. Phylogenomic studies have therefore been manual or semiautomated. In addition to taking considerable user time, this makes phylogenomic analyses difficult to reproduce, compare, and extend. In addition, methodological improvements made in the context of one study often cannot be easily applied and evaluated in the context of other studies. We present Agalma, an automated tool that conducts phylogenomic analyses. The user provides raw Illumina transcriptome data, and Agalma produces annotated assemblies, aligned gene sequence matrices, a preliminary phylogeny, and detailed diagnostics that allow the investigator to make extensive assessments of intermediate analysis steps and the final results. Sequences from other sources, such as externally assembled genomes and transcriptomes, can also be incorporated in the analyses. Agalma tracks provenance, profiles processor and memory use, records diagnostics, manages metadata, and enables rich HTML reports for all stages of the analysis. Agalma includes a test data set and a built-in test analysis of these data. In addition to describing Agalma, we here present a sample analysis of a larger seven-taxon data set. Agalma is available for download at https://bitbucket.org/caseywdunn/agalma. Agalma allows complex phylogenomic analyses to be implemented and described unambiguously as a series of high-level commands. This will enable phylogenomic studies to be readily reproduced, modified, and extended.

Background

Phylogenomic studies require many coordinated steps, yet integrated tools to conduct them are lacking. Manual and semiautomated analyses are technically demanding and difficult to reproduce, compare, and extend.

Background

  • Phylogenomic analyses span many matrix-construction steps, from raw-read filtering and assembly through homology and orthology identification, alignment, inference, and result summarization.They also require cross-step reporting, records, file reformatting, and balancing computational load.Source: Background · p. 2
  • Manual or semiautomated analyses demand extensive technical effort to reproduce and can make evaluation, cross-study integration, expansion, and comparison of alternatives difficult.Manual analyses also involve subjective decisions that may affect final results.Source: Background · p. 2
  • Existing tools addressed subsets of the workflow, but no existing tool could execute a full phylogenomic analysis.PartiGene aids assembly and annotation of Sanger transcriptome data, while SCaFoS supports curation of supermatrices from assembled transcriptomes.Source: Background · pp. 2–3
  • Agalma automates analyses from raw sequence reads to preliminary trees, integrates diagnostics across datasets and steps, and supports iterative optimization before external inference.Investigators can inspect reports, update settings, and regenerate matrices and preliminary trees.Source: Background · p. 3

Implementation

Agalma combines data cataloging and modular pipelines for transcriptome assembly, gene loading, and phylogenetic analysis, with reports and test data supporting use and development.

Implementation

  • BioLite provides Agalma’s pipeline framework for provenance tracking, resource profiling, and centralized diagnostics, while cataloging records sample metadata and data paths.Raw reads and metadata can be imported from NCBI SRA and converted to FASTQ.Source: Implementation · p. 3

Assembly

  • Assembly workflows filter reads, remove rRNA, estimate insert sizes, assemble read sets under varied protocols, and annotate transcripts for contaminants, splice variants, coverage, and protein similarity.The pipeline can assemble multiple read subsets, supporting comparison across sequencing effort.Source: Assembly · pp. 4–5
  • Agalma’s Trinity wrapper filters undersized components before Butterfly and uses GNU parallel for more efficient load balancing than ParaFly.For five test-data assemblies, the filter reduces Butterfly commands from roughly 100,000 to 60,000.Source: Assembly · p. 5

Load genes into the local Agalma database

  • The load command adds assembled datasets and externally sourced gene predictions to the local database for downstream phylogenetic analysis.It is typically run once per assembled dataset and once per set of external gene predictions.Source: Load genes into the local Agalma… · p. 5

Phylogenetic analysis

  • Phylogenetic pipelines cluster homologs, align and clean sequences, infer gene trees, and prune trees into maximally inclusive subtrees with at most one sequence per taxon.The resulting putative orthologs can support a supermatrix and preliminary maximum-likelihood species tree, with further inference available through external tools.Source: Phylogenetic analysis · p. 6

Test data and analyses

  • The bundled test data support installation checks, regression testing, examples, and feature development, using a small dataset chosen to run relatively quickly.They include read pairs, gene sequences from five siphonophores, and gene predictions from a genome assembly.Source: Test data and analyses · p. 7

Results and Discussion

In a seven-species example, Agalma generated a preliminary species tree congruent with previous siphonophore analyses and reports that expose assembly, gene-retention, and resource-use diagnostics. Public data and runnable scripts illustrate how the analysis can be repeated and adapted.

Results and Discussion

  • The seven-species example combines new Illumina reads for five siphonophores with gene predictions for two outgroups, and its deposited data and scripts can repeat or adapt the analysis.The scripts generate the analysis reports, and the figures are taken from those reports.Source: Results and Discussion · pp. 7–8

Assembly

  • Assembly reports show rRNA fractions ranging from 0.4% to 27.2% and average insert sizes of 266 bp across the example samples.Detailed reports provide diagnostics on sample quality and library preparation.Source: Assembly · p. 8

Phylogenetic analyses

  • The preliminary species tree was congruent with previous analyses of siphonophore relationships.Cluster refinement removed the most genes, largely because of taxon-sampling criteria; treeprune caused the next major reduction through orthology uncertainty and poor sampling.Source: Phylogenetic analyses · p. 8

Resource Utilization

  • The transcriptome assembly step used the most memory and took the longest, while resource plots report time and maximum memory for external executables.These profiles support resource planning, computational-load balancing, and analysis optimization.Source: Resource Utilization · p. 8

Conclusions

Agalma combines automated phylogenomic analysis with detailed diagnostics, supporting close inspection while reducing the effort of analysis. Its workflow also supports reproducibility and comparison of alternative methods.

Conclusions

  • Detailed diagnostics let investigators inspect results closely while automation provides efficiencies that leave more time for assessment.The paper also says improvements made in one study can be applied to other studies.Source: Conclusions · p. 9
  • Agalma enables exact reproduction of an analysis from its code, without readers or reviewers re-curating the dataset or rewriting the analysis code.The paper states that reproducibility requires data and analysis to be described explicitly and unambiguously.Source: Conclusions · p. 9
  • Agalma can be expanded to include alternative methods and compare them within a complete workflow consistent in other respects.The paper gives ortholog-identification methods as an example of alternatives.Source: Conclusions · p. 9

Availability and requirements

Agalma is listed as a Python program for Linux and Mac OS X, with BioLite as an additional requirement.

Availability and requirements

Author’s contributions

CWD, MH, and FZ shared design, implementation, testing, and manuscript work, with MH doing most of the coding and MH and FZ conducting the analyses.

Author’s contributions

Figures

The figures show transcriptome assembly diagnostics, gene-sequence counts across matrix construction, the preliminary phylogeny, and computational resource use.

Figures

  • Figure 1 reports sequencing-read distribution across genes and assembled-gene size distributions, including a comparison of all genes with those having a Swiss-Prot hit.Source: Figures · p. 13, Figure 1 · p. 13
  • Most gene sequences are eliminated during the first homology-evaluation step, with many of the remaining sequences removed during orthology evaluation.Source: Figures · p. 14, Figure 2 · p. 14
  • Figure 3 presents the preliminary maximum-likelihood phylogeny inferred from a protein supermatrix under the WAG + Γ model.Source: Figure 3 · p. 15, Figures · p. 15
  • Figure 4 profiles time and maximum memory use by external executables; assembly is the transcriptome pipeline's peak-memory and longest-running step.Source: Figure 4 · p. 16

Additional Files

The additional files provide HTML reports for assembly of the test data sets and phylogenetic analysis of the sample data.

Additional file 1 — HTML report for assembly of the sample data sets.

  • The assembly report summarizes five ingroup assemblies and links to detailed species-level reports through the Catalog ID column.Annotated-transcript FASTA files were removed to reduce file size.Source: Additional file 1 — HTML report for… · p. 17

Additional file 2 — HTML report for phylogenetic analyses.

Loading 1307.6432v1…