-->PONE-D-26-09692-->-->What topological and geometric structure do biological foundation models learn? Evidence from 141 hypotheses-->-->PLOS One

Dear Dr. Kendiukhov,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Please submit your revised manuscript by Jun 19 2026 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:-->

• A letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.
• A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.
• An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

-->

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

As the corresponding author, your ORCID iD is verified in the submission system and will appear in the published article. PLOS supports the use of ORCID, and we encourage all coauthors to register for an ORCID iD and use it as well. Please encourage your coauthors to verify their ORCID iD within the submission system before final acceptance, as unverified ORCID iDs will not appear in the published article. Only  the individual author can complete the verification step; PLOS staff cannot  verify ORCID iDs on behalf of authors.

We look forward to receiving your revised manuscript.

Kind regards,

Alexander Kolpakov

Academic Editor

PLOS One

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. Please note that PLOS One has specific guidelines on code sharing for submissions in which author-generated code underpins the findings in the manuscript. In these cases, we expect all author-generated code to be made available without restrictions upon publication of the work. Please review our guidelines at https://journals.plos.org/plosone/s/materials-and-software-sharing#loc-sharing-code and ensure that your code is shared in a way that follows best practice and facilitates reproducibility and reuse.

3. Please ensure that you refer to Figure 2, 3, 4, 5, 6, 8 in your text as, if accepted, production will need this reference to link the reader to the figure.

4. If the reviewer comments include a recommendation to cite specific previously published works, please review and evaluate these publications to determine whether they are relevant and should be cited. There is no requirement to cite these works unless the editor has indicated otherwise.

5. Please review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the rebuttal letter that accompanies your revised manuscript. If you need to cite a retracted article, indicate the article’s retracted status in the References list and also include a citation and full reference for the retraction notice.

Additional Editor Comments :

It can be accepted after the minor revision recommended by the referee.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

-->Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented. -->

Reviewer #1: Partly

**********

-->2. Has the statistical analysis been performed appropriately and rigorously? -->

Reviewer #1: No

**********

-->3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.-->

Reviewer #1: Yes

**********

-->4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.-->

Reviewer #1: No

**********

-->5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)-->

Reviewer #1: 1. Summary of the Manuscript

The manuscript entitled “What topological and geometric structure do biological foundation models learn? Evidence from 141 hypotheses” investigates whether biological foundation models such as scGPT and Geneformer encode meaningful geometric and topological structures within their internal representations. The study employs an autonomous hypothesis-generation and testing framework to evaluate 141 hypotheses across multiple experimental conditions, datasets, and null models. The authors aim to distinguish genuine biological structure from statistical artifacts through rigorous validation and extensive null-model comparisons.

The work lies at the intersection of mechanistic interpretability, computational biology, and topological data analysis, addressing an important question regarding the interpretability and reliability of modern biological foundation models.

2. Strong Aspects

2.1 Innovativeness and Conceptual Contribution

The manuscript demonstrates a high degree of innovation, particularly through its use of an autonomous executor–brainstormer loop to systematically generate and evaluate hypotheses. This approach represents a significant methodological advancement compared to traditional hypothesis-driven research, as it allows exploration of a vast hypothesis space while explicitly documenting both positive and negative results.

Furthermore, the integration of advanced analytical techniques—such as persistent homology, manifold distance metrics, canonical correlation analysis, and community detection—reflects a sophisticated and interdisciplinary approach. The combination of topological data analysis with biological foundation models is especially noteworthy, as it extends interpretability research beyond linear geometric representations into nonlinear and higher-order structures.

2.2 Methodological Rigor and Data Analysis

The study is methodologically robust and carefully structured. Key strengths include:

The use of multiple datasets and tissue domains (immune, lung, and external-lung), which supports cross-domain validation.

Implementation of disjoint gene-pool splits to prevent information leakage, enhancing the reliability of generalization claims.

A hierarchical null-model framework, ranging from feature shuffling to strict max-null auditing, which provides a rigorous benchmark for distinguishing real signal from artifacts.

The inclusion of multiple evaluation metrics such as AUROC improvements, null-gap analysis, and domain-split pass rates further strengthens the analytical depth of the study.

2.3 Performance and Empirical Findings

The results demonstrate that the proposed analytical framework successfully identifies meaningful geometric and topological structures:

Strong cross-model alignment (canonical correlation ≈ 0.80) indicates that independently trained models converge on similar geometric representations.

Persistent homology analysis reveals statistically significant topological features across most transformer layers, suggesting non-trivial structural organization in embedding spaces.

The identification of a hierarchy of distance metrics, where manifold-based distances outperform Euclidean distance, provides valuable insight into how biological relationships are encoded.

Additionally, the discovery that regulatory motifs align with geometric community structure represents a particularly strong and biologically meaningful finding.

2.4 Practical Relevance and Impact

The study has substantial implications for both computational biology and machine learning:

It advances understanding of how foundation models encode biological knowledge, which is critical for trust and interpretability.

The findings suggest potential applications in gene regulatory network inference, feature selection, and biological discovery.

The systematic mapping of both positive and negative results contributes to reducing publication bias and improving reproducibility in interpretability research.

Overall, the work bridges theoretical modeling and practical biological insight, making it highly relevant for real-world applications.

3. Weak Aspects

3.1 Dataset Scope and Generalizability

Despite the use of multiple tissue domains, the study relies on a limited number of datasets and focuses primarily on specific biological contexts. The results show that strong signals are concentrated in immune tissue, while lung and external-lung domains exhibit weaker or fragile effects under strict validation.

This raises concerns about the generalizability of the findings across broader biological systems. The observed domain dependency suggests that conclusions may not extend uniformly to other tissues or datasets.

3.2 Validation Complexity and Interpretability

While the hierarchical null-model framework is a strength, it also introduces complexity that may hinder interpretability. The large number of hypotheses (141) and multiple layers of validation make it challenging to clearly isolate the contribution of individual methodological components.

Moreover, although negative results are documented, the manuscript could provide a clearer synthesis of how these failures inform the overall theoretical conclusions.

3.3 Overfitting and Dependence on External Annotations

There is evidence that some performance improvements may be partially driven by confounding factors:

The reduction in performance under coexpression-matched null models suggests that some geometric signals may reflect known biological correlations rather than novel insights.

The strongest results (e.g., motif–community alignment) rely on external regulatory annotations, indicating that the model does not independently encode all relevant biological information.

Additionally, the degradation of robustness when incorporating additional biological annotations highlights potential overfitting or correlation with null structures.

3.4 Missing Ethical, Practical, and Scalability Considerations

The manuscript does not sufficiently address broader implications:

Ethical considerations, such as biases in biological datasets or downstream decision-making risks, are not discussed.

Computational cost and scalability of the hypothesis-screening framework are not evaluated, despite its iterative and resource-intensive nature.

Practical deployment challenges, including integration into biological pipelines or clinical applications, are not explored.

These omissions limit the applicability of the study in real-world contexts.

4. Recommended Changes

4.1 Expand Dataset Diversity and Validation Scope

To improve generalizability, the authors should:

Include additional datasets covering diverse biological conditions and species.

Evaluate the method on independent external benchmarks not used during hypothesis generation.

Provide a more detailed analysis of domain-specific variability, particularly why immune tissue shows stronger signals.

4.2 Strengthen Validation and Interpretability

The manuscript would benefit from:

Clearer organization of hypothesis families and their relative contributions.

Additional statistical validation, including significance testing across multiple comparisons.

Ablation studies to isolate the impact of key components (e.g., topology vs. community structure).

4.3 Improve Robustness and Reduce Overfitting Risks

To enhance methodological robustness, the authors are encouraged to:

Apply cross-validation and systematic hyperparameter tuning across experiments.

Incorporate regularization strategies or noise-based validation to test stability.

Further analyze the impact of confounding variables such as gene coexpression.

4.4 Address Ethical and Practical Considerations

The manuscript should include:

A discussion on ethical implications, including potential biases in biological data and interpretability risks.

Evaluation of computational complexity and scalability of the proposed framework.

Consideration of how the approach can be integrated into practical biological or biomedical workflows.

suggested citation: https://doi.org/10.1038/s41598-026-39632-y https://doi.org/10.1016/j.jnca.2026.104462 https://doi.org/10.1007/s10462-026-11504-x https://doi.org/10.1007/s13369-025-10647-3

**********

-->6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review?  For information about this choice, including consent withdrawal, please see our Privacy Policy.-->

Reviewer #1: No

**********

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

To ensure your figures meet our technical requirements, please review our figure guidelines: https://journals.plos.org/plosone/s/figures

You may also use PLOS’s free figure tool, NAAS, to help you prepare publication quality figures: https://journals.plos.org/plosone/s/figures#loc-tools-for-figure-preparation.

NAAS will assess whether your figures meet our technical requirements by comparing each figure against our figure specifications.

-->

<p>What topological and geometric structure do biological foundation models learn? Evidence from 141 hypotheses

PONE-D-26-09692R1

Dear Dr. Kendiukhov,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice will be generated when your article is formally accepted. Please note, if your institution has a publishing partnership with PLOS and your article meets the relevant criteria, all or part of your publication costs will be covered. Please make sure your user information is up-to-date by logging into Editorial Manager at Editorial Manager® and clicking the ‘Update My Information' link at the top of the page. For questions related to billing, please contact billing support.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Alexander Kolpakov

Academic Editor

PLOS One

Additional Editor Comments (optional):

Reviewers' comments: