iSEEtree: interactive explorer for hierarchical data

Akewak Jeba; Ely Seraidarian; Giulio Benedetti; Leo Lahti; Theotime Pralas; Tuomas Borman

REVIEW 2 major objections 4 minor 14 references

iSEEtree: interactive explorer for hierarchical data

T0 review · 2 major / 4 minor · reviewed 2026-08-11 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read iSEEtree provides a visual interface for exploring hierarchical TreeSummarizedExperiment objects, letting users analyse microbiome data without writing R code.

desk verdict A real, installable iSEE extension for TreeSE data whose interactive integration is asserted more than demonstrated, and whose 'no programming' claim is undercut by its own Limitations section—still worth a proper referee. read the letter →

arxiv 2412.02882 v2 pith:MVZ6R5XI submitted 2024-12-03 cs.MS cs.GRq-bio.GN

Giulio Benedetti , Ely Seraidarian , Theotime Pralas , Akewak Jeba , Tuomas Borman , Leo Lahti This is my paper · ORCID

classification cs.MScs.GRq-bio.GN

keywords interactivevisualizationhierarchicaldataTreeSummarizedExperimentmicrobiomeanalysisRShinyordinationexplorationreproducibility

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper introduces iSEEtree, an R Shiny app that gives researchers a point-and-click interface for exploring TreeSummarizedExperiment objects, which store multi-table data together with hierarchical trees. The aim is to lower the programming barrier for analysing hierarchical data such as microbiome taxonomic tables. If the paper is right, users can visually inspect abundance, ordination, and tree structure without writing R code, while still inheriting iSEE's dynamic linking and reproducibility features. The authors demonstrate the app on a gut microbiome dataset and argue that the same generic panels apply to any hierarchical multi-table dataset.

What carries the argument

The central object is the TreeSummarizedExperiment (TreeSE) container, an S4 class that extends SummarizedExperiment and SingleCellExperiment with rowTree, colTree, rowLinks, and colLinks elements to represent hierarchical relations among features and samples. The app's interface is built on the iSEE panel framework, which provides dynamic linking between panels, selectable parameters, and code tracking. iSEEtree adds hierarchical-aware panels to that framework, so the machinery carrying the argument is the combination of a standardized hierarchical data container with a proven interactive plotting engine.

What would settle it

Open a TreeSE object with a large row tree in iSEEtree, select one clade in the Row tree panel, and check whether the linked abundance and heatmap panels update exactly to the members of that clade; if the selection mismatches or the app freezes, the claim that it provides reliable interactive hierarchical exploration fails.

Watch

Extended reading notes

Core claim

iSEEtree is an R Shiny application that extends the interactive exploration model of iSEE to TreeSummarizedExperiment objects. It adds custom panels for abundance barplots and density plots, reduced-dimension and RDA plots, loading plots, and row and column tree visualisations, all connected through iSEE's dynamic selection system. The paper reports that the app inherits the full iSEE feature set, including linked panels, code tracking for reproducibility, interactive tours, custom colouring, and downloadable plots, and demonstrates the workflow on a microbiome dataset derived from Tengeler et al. (2020). Its central contribution is a general-purpose, graphical front end for hierarchical data stored in the TreeSE container.

Load-bearing premise

The central assumption is that iSEE's interactive panel machinery works reliably when extended to TreeSummarizedExperiment objects, so that linked selections and visualisations behave correctly on hierarchical data.

Editorial extensions

If this is right

Users can explore abundance, ordination, loading, and tree views side by side, with selections in one panel dynamically linked to the others.
The generated R scripts from an interactive session support reproducible reporting, turning clicks into documented analysis steps.
Because the app is built on the generic TreeSE container, the same panels should work for any hierarchical multi-table dataset, not only microbiome count tables.
The paper reports that iSEEtree already underpins a fully graphical web app for importing, manipulating, and visualising microbiome data, extending the same interface to users who do not use R directly.
Novice R users can perform compositional and ordination exploration without learning the underlying plotting commands.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The paper does not test this, but the same panel framework could plausibly be applied to other TreeSE use cases, such as single-cell lineage trees or gene-family trees in comparative genomics, wherever a hierarchical grouping of rows or columns exists.
Because the limitations section names scalability as a concern, a natural stress test would compare iSEEtree's response time with command-line plotting on the same large TreeSE object to quantify the interaction overhead.
The code-tracking feature may lower the barrier to reproducible analysis for beginners, turning clicks into documented scripts that can later be edited by hand, though the paper does not evaluate this educational effect.
If the interface were extended to support constructing TreeSE objects directly, as the limitations suggest, the tool would become accessible to an even broader audience than the current R-based launch workflow allows.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

You should know this is a genuine software contribution, not vaporware. iSEEtree is on Bioconductor, the code is real, and the paper shows it adding hierarchical panels (Row tree, Column tree, Abundance) to the iSEE framework for TreeSummarizedExperiment objects. That combination is new, and for tool users in the microbiome/Bioconductor world it's a reasonable niche to fill. The paper describes the app clearly, documents its relationship to mia and iSEE, and demonstrates it on a real dataset. Credit where due: it's openly licensed, has online documentation, and is part of a larger ecosystem (miaDash) that actually uses it. No equations to check, but no invented claims either—the artifact is reproducible, and the citation pattern is honest, mostly to the packages it builds on.

The soft spots are real but not fatal. The abstract says users can explore 'without the need for extensive knowledge of R programming,' and the paper repeats that claim. Then the Limitations section says you still need to build TreeSE objects and launch the app from R or RStudio. That's a direct contradiction, and it should be fixed by toning down the 'no programming' language and instead saying 'lowered barrier' or 'no plotting code required after your TreeSE is ready.' Minor wording fix, but the paper's own text makes it hard to defend the headline.

The second, more substantive gap is the dynamic linking claim. The paper says iSEEtree 'inherits the complete array' of iSEE functionality, including dynamic linking, tours, code tracking, for the new panels. But no demonstration or test shows that the new tree and abundance panels actually participate in dynamic linking or emit reproducible code. The only evidence is a static screenshot. This is the central interactive claim, so it's worth probing. My guess is it probably works—the panels are built as iSEE panel classes, and iSEE's architecture is designed for that—but the paper doesn't prove it. An editor could reasonably ask for a short supplementary video, a worked example showing linked selections, or a vignette that exercises the code-tracking feature on the new panels. That would close the gap without requiring any new science.

Minor quibble: no performance benchmarks, but the Limitations section acknowledges scaling limits, so that's consistent. The paper is a typical application note: modest scope, honest about constraints except for the one overclaim.

Bottom line: I'd send it to peer review. It's a real tool that a chunk of the community will use, and the issues are addressable with edits and a small verification step. My own verdict would be 'conditional accept' with the usability language fixed and the interactive claim demonstrated. Worth a reading group if anyone in your group works on microbiome viz or Shiny-based tools.

Referee Report

2 major / 4 minor

Summary. The paper presents iSEEtree, an R Shiny application built on the iSEE package that provides a graphical interface for exploring TreeSummarizedExperiment (TreeSE) objects. The central claim is that iSEEtree extends iSEE's interactive graphics to hierarchical data structures, enabling users to explore microbiome and other multi-table datasets through panels such as Abundance, Reduced dimension, Loading, Row tree, and Column tree without writing plotting code. The authors describe the workflow (import, construct TreeSE, launch app), list the panel types, and illustrate the interface with a screenshot from a mouse gut microbiome dataset (Tengeler et al., 2020). The paper also discusses related work, the TreeSE container, integration with the mia ecosystem (e.g., miaDash), and three limitations: scalability, restricted panel coverage, and the need to build TreeSE objects and launch the app from R or RStudio.

Significance. If the central claim holds, iSEEtree is a valuable contribution to the Bioconductor ecosystem because it lowers the technical barrier for exploring hierarchical data, particularly in microbiome research, and it provides a foundation for domain-specific graphical tools such as miaDash. The software is openly available on Bioconductor with documentation and a tutorial, which is a concrete strength. The paper convincingly argues that adopting the TreeSE container gives the tool broader applicability than phyloseq-based solutions. However, the paper overstates its case in two ways that affect the strength of the contribution: the abstract's 'without the need for extensive knowledge of R programming' is in tension with the Limitations section, and the claim that the new TreeSE-specific panels inherit iSEE's dynamic linking, code tracking, and tours is asserted but not demonstrated. These are fixable issues, but they are central to the paper's stated motivation.

major comments (2)

[Software Implementation] The paragraph beginning 'Consequently, our app inherits the complete array...' asserts that the new TreeSE-specific panels (Abundance, Row tree, Column tree) inherit iSEE's dynamic linking, code tracking, panel tours, and downloadable results. This is load-bearing for the paper's central claim that iSEEtree 'expands the interactive graphics capabilities of related work to hierarchical structures.' Yet the manuscript provides no evidence that these interactive features actually work for the new panels: the only demonstration is a static screenshot (Fig. 1B), and the Limitations section does not address whether dynamic linking or code tracking function for hierarchical panels. If these features do not work, the app would be a collection of static plots rather than an interactive explorer. I ask the authors to provide a concrete demonstration—for example, a short reproducible example or animation showing a selection in the Row tree panel being reflected in an Abundance or Reduced dimension panel, and the corresponding R code generated by the 'code tracking' feature. Without such verification, the central claim is not yet established.
[Abstract and Limitations] The abstract and introduction state that users can 'interactively explore several aspects of their data without the need for extensive knowledge of R programming,' but the Limitations section says that 'it is still necessary to build TreeSE objects and launch the app either from R or RStudio' and that this 'may represent a barrier for novice users.' These statements are in tension: constructing TreeSE objects typically requires R programming, and launching from R/RStudio itself requires some R familiarity. The authors should either clarify that the intended user is someone who already works with R/Bioconductor and only needs to avoid writing plotting code, or soften the 'without the need' claim to 'with reduced need' or 'with minimal R knowledge.' As written, the motivation is undercut by the paper's own limitation.

minor comments (4)

[Operative framework] The text says the app is 'launched from an R console or RStudio by executing the command iSEE(tse),' but the package is named iSEEtree; please clarify whether this is a re-export of iSEE's function or a new function with the same name, to avoid confusion for users.
[Applications / Use cases] The phrase '16rRNA gene sequencing' should be corrected to '16S rRNA gene sequencing' for standard terminology.
[Figure 1] In Fig. 1B, the panel labels (i) through (vi) are referred to in the caption but the caption does not state which panel type corresponds to each label; adding explicit panel names (e.g., 'Abundance', 'Row tree', 'Reduced dimension') would make the figure self-contained.
[Software Implementation] The sentence 'However, certain panels require the presence of specific elements in the TreeSE object, and they will appear only when these are available' would benefit from a concrete example (e.g., RDA plot requires a reducedDim named 'RDA'), so that users know what to expect.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the central claim is software functionality, not a derived result; dependencies are independently built and disclosed.

full rationale

The paper is an applications note describing software, not a derivation with equations, fitted parameters, or quantitative predictions. The central claim, that iSEEtree provides a visual interface for exploring TreeSummarizedExperiment objects, is a functional statement about an R/Shiny application, supported by package availability, documentation, and a demonstration dataset. The stated dependence on earlier packages (iSEE, mia, TreeSE) is an explicit disclosure of implementation reuse, not a circular derivation: iSEE's interactive panel framework is an independently developed, externally available Bioconductor package, and iSEEtree's new TreeSE-specific panels are presented as an extension rather than as a prediction forced by data. The skeptical concern that dynamic linking and code tracking are asserted but not demonstrated in the paper is a validation or completeness gap, not circularity, because the assertion concerns inherited software behavior, not a result derived from inputs. Self-citations (mia, miaDash, Orchestrating Microbiome Analysis) serve to situate the software in its ecosystem and to point to existing tooling; they do not make the central claim true by construction. No self-definitional, fitted-input-as-prediction, uniqueness-imported, ansatz-smuggled, or renamed-known-result pattern is present. The derivation chain, such as it is, is self-contained as a software description rather than as an analytical proof.

Assumptions & free parameters 0 free parameters · 2 assumptions · 0 invented entities

This is a software application note; the central claim is about the existence and utility of a tool, not about a scientific phenomenon. No numerical parameters are fitted, no axioms are proved, and no new physical or biological entities are postulated. The only assumptions are about the soundness and generality of the underlying software frameworks.

assumptions (2)

domain assumption The TreeSummarizedExperiment container can represent hierarchical multi-table data in a general way across domains.
Invoked throughout the paper, especially in the 'Data containers' section, as the basis for iSEEtree's claimed generality to 'a vast array of datasets with a hierarchical structure.'
domain assumption The iSEE package's interactive panel framework is stable and supports the new hierarchical panel types added by iSEEtree.
The 'Software Implementation' and 'App functionality' sections state that iSEEtree 'derives its graphical and interactive capabilities from the mia and iSEE packages' and inherits features such as dynamic linking. If this extension point did not work, the app would not deliver its promised functionality.

how reviews work

0 comments

Cite this review

Pith. "Pith review of iSEEtree: interactive explorer for hierarchical data." pith.science (2026). https://pith.science/paper/MVZ6R5XI

@misc{pith2026241202882,
  author       = {Pith},
  title        = {Pith review of: iSEEtree: interactive explorer for hierarchical data},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MVZ6R5XI}},
  note         = {Machine review of arXiv:2412.02882}
}

abstract

$\textbf{Motivation:}$ Hierarchical data structures are prevalent across several fields of research, as they represent an organised and efficient approach to study complex interconnected systems. Their significance is particularly evident in microbiome analysis, where microbial communities are classified at various taxonomic levels along the phylogenetic tree. In light of this trend, the R/Bioconductor community has established a reproducible analytical framework for hierarchical data, which relies on the highly generic and optimised TreeSummarizedExperiment data container. However, using this framework requires basic proficiency in programming. $\textbf{Results:}$ To reduce the entry requirements, we developed iSEEtree, an R shiny app which provides a visual interface for the analysis and exploration of TreeSummarizedExperiment objects, thereby expanding the interactive graphics capabilities of related work to hierarchical structures. This way, users can interactively explore several aspects of their data without the need for extensive knowledge of R programming. We describe how iSEEtree enables the exploration of hierarchical multi-table data and demonstrate its functionality with applications to microbiome analysis. $\textbf{Availability and Implementation:}$ iSEEtree was implemented in the R programming language and is available on Bioconductor at https://bioconductor.org/packages/iSEEtree under an Artistic 2.0 license. $\textbf{Contact:}$ giulio.benedetti@utu.fi or leo.lahti@utu.fi.

Figures

Figures reproduced from arXiv: 2412.02882 by the authors.

**Figure 1.** iSEEtree operative framework. A, Analytical workflow. (i) Data in standardised format is imported in R. This may include abundance assays and optional metadata on features, samples and tree hierarchies. (ii) A TreeSE object is constructed from the data. This is an optimised and standardised container for hierarchical data analysis. (iii) The object is input to iSEEtree, which launches an interface with a set of data… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

14 extracted references · 12 canonical work pages

[1]

Orchestrating single-cell analysis with bioconductor

Robert A Amezquita, Aaron TL Lun, Etienne Becht, Vince J Carey, Lindsay N Carpp, Ludwig Geistlinger, Federico Marini, Kevin Rue-Albrecht, Davide Risso, Charlotte Soneson, et al. Orchestrating single-cell analysis with bioconductor. Nature methods, 17 0 (2): 0 137--145, 2020. doi:10.1038/s41592-019-0654-x

work page doi:10.1038/s41592-019-0654-x 2020
[2]

miaDash: Shiny app for interactive analysis and exploration of microbiome data, 2024

Giulio Benedetti and Leo Lahti. miaDash: Shiny app for interactive analysis and exploration of microbiome data, 2024. URL https://github.com/microbiome/miaDash. R package version 0.99.2

work page 2024
[3]

Ernst, Sudarshan A

Tuomas Borman, Felix G.M. Ernst, Sudarshan A. Shetty, and Leo Lahti. mia: Microbiome analysis, 2024. URL https://bioconductor.org/packages/mia. R package version 1.14.0

work page 2024
[4]

shiny: Web Application Framework for R, 2024

Winston Chang, Joe Cheng, JJ Allaire, Carson Sievert, Barret Schloerke, Yihui Xie, Jeff Allen, Jonathan McPherson, Alan Dipert, and Barbara Borges. shiny: Web Application Framework for R, 2024. URL https://CRAN.R-project.org/package=shiny. R package version 1.9.1

work page 2024
[5]

Bioconductor: open software development for computational biology and bioinformatics

Robert C Gentleman, Vincent J Carey, Douglas M Bates, Ben Bolstad, Marcel Dettling, Sandrine Dudoit, Byron Ellis, Laurent Gautier, Yongchao Ge, Jeff Gentry, et al. Bioconductor: open software development for computational biology and bioinformatics. Genome biology, 5: 0 1--16, 2004. doi:10.1186/gb-2004-5-10-r80

doi:10.1186/gb-2004-5-10-r80 2004
[6]

Microbiome datasets are compositional: and this is not optional

Gregory B Gloor, Jean M Macklaim, Vera Pawlowsky-Glahn, and Juan J Egozcue. Microbiome datasets are compositional: and this is not optional. Frontiers in microbiology, 8: 0 2224, 2017. doi:10.3389/fmicb.2017.02224

work page arXiv 2017
[7]

Treesummarizedexperiment: a s4 class for data with hierarchical structure

Ruizhu Huang, Charlotte Soneson, Felix GM Ernst, Kevin C Rue-Albrecht, Guangchuang Yu, Stephanie C Hicks, and Mark D Robinson. Treesummarizedexperiment: a s4 class for data with hierarchical structure. F1000Research, 9, 2020. doi:10.12688/f1000research.26669.2

work page doi:10.12688/f1000research.26669.2 2020
[8]

Orchestrating microbiome analysis with bioconductor [beta version], 2021

L Lahti, S Sudarshan, FM Ernst, et al. Orchestrating microbiome analysis with bioconductor [beta version], 2021

work page 2021

Show all 14 references

[9]

Shiny-phyloseq: Web application for interactive microbiome analysis with provenance tracking

Paul J McMurdie and Susan Holmes. Shiny-phyloseq: Web application for interactive microbiome analysis with provenance tracking. Bioinformatics, 31 0 (2): 0 282--283, 2015. doi:10.1093/bioinformatics/btu616

2015 doi
[10]

SummarizedExperiment: SummarizedExperiment container, 2024

Martin Morgan, Valerie Obenchain, Jim Hester, and Hervé Pagès. SummarizedExperiment: SummarizedExperiment container, 2024. URL https://bioconductor.org/packages/SummarizedExperiment. R package version 1.35.2

2024
[11]

R: A Language and Environment for Statistical Computing

R Core Team . R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria, 2024. URL https://www.R-project.org/

2024
[12]

isee: interactive summarizedexperiment explorer

Kevin Rue-Albrecht, Federico Marini, Charlotte Soneson, and Aaron TL Lun. isee: interactive summarizedexperiment explorer. F1000Research, 7, 2018. doi:10.12688/f1000research.14966.1

2018 doi
[13]

Gut microbiota from persons with attention-deficit/hyperactivity disorder affects the brain in mice

Anouk C Tengeler, Sarita A Dam, Maximilian Wiesmann, Jilly Naaijen, Miranda Van Bodegom, Clara Belzer, Pieter J Dederen, Vivienne Verweij, Barbara Franke, Tamas Kozicz, et al. Gut microbiota from persons with attention-deficit/hyperactivity disorder affects the brain in mice. ...

2020 doi
[14]

Evan Johnson

Yue Zhao, Anthony Federico, Tyler Faits, Solaiappan Manimaran, Daniel Segrè, Stefano Monti, and W. Evan Johnson. animalcules: interactive microbiome analytics and visualization in r. Microbiome, 9: 0 76, 2021. doi:10.1186/s40168-021-01013-0

2021 doi

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.

[1] [1]

Orchestrating single-cell analysis with bioconductor

Robert A Amezquita, Aaron TL Lun, Etienne Becht, Vince J Carey, Lindsay N Carpp, Ludwig Geistlinger, Federico Marini, Kevin Rue-Albrecht, Davide Risso, Charlotte Soneson, et al. Orchestrating single-cell analysis with bioconductor. Nature methods, 17 0 (2): 0 137--145, 2020. doi:10.1038/s41592-019-0654-x

work page doi:10.1038/s41592-019-0654-x 2020

[2] [2]

miaDash: Shiny app for interactive analysis and exploration of microbiome data, 2024

Giulio Benedetti and Leo Lahti. miaDash: Shiny app for interactive analysis and exploration of microbiome data, 2024. URL https://github.com/microbiome/miaDash. R package version 0.99.2

work page 2024

[3] [3]

Ernst, Sudarshan A

Tuomas Borman, Felix G.M. Ernst, Sudarshan A. Shetty, and Leo Lahti. mia: Microbiome analysis, 2024. URL https://bioconductor.org/packages/mia. R package version 1.14.0

work page 2024

[4] [4]

shiny: Web Application Framework for R, 2024

Winston Chang, Joe Cheng, JJ Allaire, Carson Sievert, Barret Schloerke, Yihui Xie, Jeff Allen, Jonathan McPherson, Alan Dipert, and Barbara Borges. shiny: Web Application Framework for R, 2024. URL https://CRAN.R-project.org/package=shiny. R package version 1.9.1

work page 2024

[5] [5]

Bioconductor: open software development for computational biology and bioinformatics

Robert C Gentleman, Vincent J Carey, Douglas M Bates, Ben Bolstad, Marcel Dettling, Sandrine Dudoit, Byron Ellis, Laurent Gautier, Yongchao Ge, Jeff Gentry, et al. Bioconductor: open software development for computational biology and bioinformatics. Genome biology, 5: 0 1--16, 2004. doi:10.1186/gb-2004-5-10-r80

doi:10.1186/gb-2004-5-10-r80 2004

[6] [6]

Microbiome datasets are compositional: and this is not optional

Gregory B Gloor, Jean M Macklaim, Vera Pawlowsky-Glahn, and Juan J Egozcue. Microbiome datasets are compositional: and this is not optional. Frontiers in microbiology, 8: 0 2224, 2017. doi:10.3389/fmicb.2017.02224

work page arXiv 2017

[7] [7]

Treesummarizedexperiment: a s4 class for data with hierarchical structure

Ruizhu Huang, Charlotte Soneson, Felix GM Ernst, Kevin C Rue-Albrecht, Guangchuang Yu, Stephanie C Hicks, and Mark D Robinson. Treesummarizedexperiment: a s4 class for data with hierarchical structure. F1000Research, 9, 2020. doi:10.12688/f1000research.26669.2

work page doi:10.12688/f1000research.26669.2 2020

[8] [8]

Orchestrating microbiome analysis with bioconductor [beta version], 2021

L Lahti, S Sudarshan, FM Ernst, et al. Orchestrating microbiome analysis with bioconductor [beta version], 2021

work page 2021