Dear Dr. Okuno,

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 30 2025 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.

• A rebuttal 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 .

We look forward to receiving your revised manuscript.

Kind regards,

Zeheng Wang

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. We note that the grant information you provided in the ‘Funding Information’ and ‘Financial Disclosure’ sections do not match.

When you resubmit, please ensure that you provide the correct grant numbers for the awards you received for your study in the ‘Funding Information’ section.

4. Thank you for stating in your Funding Statement:

[This work was supported by JST, Center of Innovation Program (JPMJCE1302).].

Please provide an amended statement that declares *all* the funding or sources of support (whether external or internal to your organization) received during this study, as detailed online in our guide for authors at http://journals.plos.org/plosone/s/submit-now. Please also include the statement “There was no additional external funding received for this study.” in your updated Funding Statement.

Please include your amended Funding Statement within your cover letter. We will change the online submission form on your behalf.

5. Thank you for stating the following in the Acknowledgments Section of your manuscript:

[This work was supported by JST, Center of Innovation Program (JPMJCE1302). This research was supported by JST Moonshot R\&D Grant Number JPMJMS2021 and JPMJMS2024.]

We note that you have provided funding information that is currently declared in your Funding Statement. However, funding information should not appear in the Acknowledgments section or other areas of your manuscript. We will only publish funding information present in the Funding Statement section of the online submission form.

Please remove any funding-related text from the manuscript and let us know how you would like to update your Funding Statement. Currently, your Funding Statement reads as follows:

[This work was supported by JST, Center of Innovation Program (JPMJCE1302).].

Please include your amended statements within your cover letter; we will change the online submission form on your behalf.

6. In the online submission form, you indicated that [Data cannot be shared publicly because of patient privacy in electronic medical records. Data are available from the corresponding author and Kyoto University Graduate School and Faculty of Medicine, Ethics Committee via email (ethcom@kuhp.kyoto-u.ac.jp) or telephone (+81-75-753-4680) for researchers who meet the criteria for access to confidential data.].

All PLOS journals now require all data underlying the findings described in their manuscript to be freely available to other researchers, either 1. In a public repository, 2. Within the manuscript itself, or 3. Uploaded as supplementary information.

This policy applies to all data except where public deposition would breach compliance with the protocol approved by your research ethics board. If your data cannot be made publicly available for ethical or legal reasons (e.g., public availability would compromise patient privacy), please explain your reasons on resubmission and your exemption request will be escalated for approval.

7. We notice that your supplementary figures are uploaded with the file type 'Figure'. Please amend the file type to 'Supporting Information'. Please ensure that each Supporting Information file has a legend listed in the manuscript after the references list.

8. Please include captions for your Supporting Information files at the end of your manuscript, and update any in-text citations to match accordingly. Please see our Supporting Information guidelines for more information: http://journals.plos.org/plosone/s/supporting-information.

Additional Editor Comments:

The manuscript presents a potentially valuable contribution to the field of AI for healthcare, particularly in its practical relevance and application-driven insights. Several reviewers acknowledged the importance of the problem being addressed and the promise of the proposed approach. However, significant concerns were raised regarding the technical presentation of the work. Specifically, the manuscript lacks the clarity, precision, and methodological rigor typically expected in AI-focused publications.

The current version suffers from a lack of domain-appropriate language and structure, which obscures the novelty and reproducibility of the methods. Multiple reviewers have recommended rejection due to these issues. Nevertheless, given the potential practical impact of the work, we have decided to offer the authors an opportunity to substantially revise and resubmit.

We strongly recommend that the authors seek assistance from a researcher or collaborator with expertise in AI or machine learning to thoroughly revise the manuscript, ensuring that the technical descriptions are accurate, the methods are transparent and reproducible, and the overall narrative meets the professional standards of the AI research community. A professional proofread focused on AI terminology and methodology is highly encouraged.

Only after a comprehensive revision addressing these concerns will the manuscript's suitability be reconsidered again for publication.

[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?

Reviewer #1: Yes

Reviewer #2: Partly

Reviewer #3: Yes

Reviewer #4: Yes

Reviewer #5: Yes

Reviewer #6: Partly

**********

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

Reviewer #1: I Don't Know

Reviewer #2: No

Reviewer #3: Yes

Reviewer #4: Yes

Reviewer #5: I Don't Know

Reviewer #6: No

**********

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

The PLOS Data policy

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: No

Reviewer #4: No

Reviewer #5: Yes

Reviewer #6: Yes

**********

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

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Yes

Reviewer #4: Yes

Reviewer #5: Yes

Reviewer #6: Yes

**********

Reviewer #1: This study introduces a deep state-space analysis framework to estimate latent patient states from electronic health records (EHR) and identify temporal risk factors for cancer progression. Overall, this framework advances interpretable deep learning for EHR analysis, offering a promising tool for personalized prognosis and treatment optimization. Addressing limitations in data accessibility and clinical complexity could amplify its impact in healthcare.

Strengths:

Innovative Methodology: Combines deep state-space models with visualization and clustering to enhance interpretability of temporal EHR data.Clinical Relevance: Identifies actionable prognostic factors (e.g., anemia, immune markers) validated on a large cancer cohort.Scalability: Framework generalizes across heterogeneous EHR data, enabling long-term disease progression modeling.

Areas for Improvement in the Manuscript:

Methodological Clarity:Provide detailed explanations of missing data handling (e.g., masking implementation) and model hyperparameter optimization processes. Clarify how the deep state-space model balances computational efficiency with accuracy, especially for large-scale EHR data.

Experimental Design: Address potential selection bias from excluding patients with <50 time steps and discuss its impact on generalizability.Acknowledge temporal shifts in medical practices or EHR recording standards (2006–2018) and their influence on data consistency.

Result Interpretation:Elaborate on the causal relationship between identified risk factors (e.g., anemia, immune markers) and disease progression, beyond correlation. Analyze heterogeneity across cancer types (Table 2) to assess whether results vary by malignancy.

Reviewer #2: The authors propose a deep state-space analysis framework that uses sequential EHR data to model and visualize patients’ latent disease states over time, enabling clustering by disease severity and identification of poor prognostic factors. The overall idea of the manuscript is nice. However, I have the following suggestions:

1. It would be better to have more consistent and precise terminology. The use of scientific and analytical language throughout the paper should be strengthened to improve clarity and help readers better grasp the study’s focus.

2. The structure of the manuscript may need to be revised. The Results and Discussion sections currently precede the Methods section. Also, part of the descriptions that belong in the Methods section can be considered relocated as part of the Results.

3. The data has been categorized several times without providing the criteria or standards. These should be explicitly defined and explained in the relevant sections.

4. The existing methods and your proposed approach should be clearly distinguished. Proper citations must be provided for existing techniques, and the manuscript should clearly state what the methodological contributions of the current work are. What is new or innovative about your approach should be made transparent.

5. Although the manuscript aims to demonstrate that deep learning is well-suited to model complex EHR data, the Methods section lacks a clear explanation of how EHR data is actually utilized in the proposed model.

6. Several descriptions in the Methods section lack clarity and precision. Please ensure the methodological steps align with the original techniques, unless improvements are introduced, in which case the modifications should be clearly stated. Terminology and phrasing must remain consistent throughout the manuscript to avoid confusion. For example, in lines 345–346, how the parameters obtained should be described clearly without omitting essential steps.

7. Are the UMAP parameters default? If parameter adjustments were made, please specify which ones and discuss their impact on the results.

8. The association between EHR test data and cluster labels (I, II, III) should be clarified.

9. The meaning of “abnormally high,” “normal,” and “abnormally low” values in each cluster should be defined within the context of this study. How is “abnormal” defined?

10. All figures should clearly label the x- and y-axes, including what each axis represents. The manuscript should include a clear description of what the figures show and their significance. Hard to correlate with text.

11. The Results and Discussion sections need more in-depth analysis. Statistical evaluations should be included to support findings, and thorough interpretations of the figures and results are necessary to enhance scientific rigor.

Reviewer #3: I enjoyed this paper and commend the authors for employing a new method to reduce dimensionality and merge temporal states and observational data. I especially appreciated the attention given to explaining the data pre-processing steps and domain knowledge used in the interpretation and use of laboratory tests. The machine-learning steps are well documented in the extended methods and are robust and supported.

Major Revision

While this paper presents a novel approach that integrates both temporal states and clinical test inputs to predict patient prognoses, the discussion section would benefit from additional supporting information and clarification:

• Your models identified anemia as a major prognostic factor. However, anemia is already a well-documented adverse effect of many of the therapies you mention, and has been widely cited as a critical clinical indicator. Did your model uncover any previously unrecognized patterns or associations involving anemia that are not well-established in the existing literature?

• How can your methods be effectively communicated to clinicians or individuals without a strong background in machine learning? What strategies or model characteristics support the trustworthiness and transparency of your predictions?

• Can you elaborate on how your methods could be adapted into clinical tools or algorithms for real-world use? For example, how might this approach be integrated into clinical decision support systems?

• Although there are understandable limitations to sharing raw patient data, could you make de-identified or simulated datasets available to support reproducibility? Your GitHub repository appears to include sample data, but this isn’t clearly highlighted—consider improving its visibility or documentation.

Minor Revision

• In Figure 1, consider making the representation of “time states” more explicit. Since temporality is a key element of your method, it would be helpful to clearly delineate different timepoints in the disease course and data collection timeline.

• Line 106: Revise to simply say “patient prognoses”—this term encompasses both favorable and unfavorable outcomes.

• Please include a brief discussion of the limitation associated with assuming survival in the absence of a recorded death in the EHR. This introduces a form of data censoring. Although most patients in your study were followed for less than two years—limiting the impact of this assumption—it may be worth suggesting future studies consider follow-up dates or discharge summaries to better estimate survival. This may be a point of confusion with clinicians who regularly employ censoring in survival analyses.

• Please check language consistency and accuracy for all figures. I found a few typos.

Reviewer #4: In this article, the authors have proposed a deep state-space analysis framework to estimate the hidden status for patients based on their observed electronic health records (EHR) data in an unsupervised fashion, to cluster the disease severity from the estimated hidden status and to identify the leading risk factors/driven biomarkers associated with each defined stratum. By accounting for the nonlinearity and time-dependency features in their framework, the deep state-space analysis method shows higher interpretability for its results compared to the conventional methods (principal component analysis, variational autoencoder, and linearized version of the proposed method). Generally speaking, this manuscript has carefully investigated the usage of deep learning methods in EHR analysis and has introduced a novel deep learning framework showing successful application in cancer research. I have several humble comments which I hope that the authors could address to further enhance the clarity and reproducibility of their research.

Major comments:

(Lines 248-254) Under either retrospective or prospective design, I have no clear idea why this proposed framework can detect causal effects in observational study as it always suffers from unmeasured confounding. I hope the authors could explain this point in more details regarding the validity of this method in the context of causal inference.

(Lines 331-333) Could you please provide more details about how the missing rate was calculated and how it is distributed at individual-level and variable-level (i.e., per-individual missing rate and per-variable missing rate). Also, I hope the authors could further explain how your proposed framework can handle large proportions of missing values.

(Table 2) In this table, the authors summarize the demographic and disease-related information for the 12,695 cancer patients in their analysis. Could you please show why the summation of the sample size for each cancer subtype is smaller than the total sample size?

(Figure 1) In the “Visualization of latent states” section, the authors mention that K-means clustering is performed before the UMAP dimension reduction (see Lines 384-386). However, Figure 1 illustrates that the UMAP step is performed prior to the clustering of latent states. What is the appropriate order of these procedures?

(Figure 2d) Please provide more details on interpreting these transition probabilities. Moreover, we observe from this figure that (1) the difference between stable -> intermediate probability and intermediate -> stable probability is 0.31% for deceased individuals and 0.12% for surviving individuals, and (2) the dangerous -> intermediate probability is larger than intermediate -> dangerous probability for deceased individuals while the dangerous -> intermediate probability is smaller than intermediate -> dangerous probability for surviving individuals. These two observations seem indicating that the deceased group is more likely to move from dangerous state to intermediate state as well as from intermediate state to stable state compared to the surviving group – could you please explain the reasons behind these observations?

Minor comments:

(Lines 327-328) In the definition of death across this study, does it mean the death due to cancer specifically, or it can arise due to other competing risks?

(Line 346) Which three networks are referred to in the phrase “the parameters of the three neural networks”?

(Lines 356-357) In the formula (1), in q distribution, x is not indexed by t – does it mean observed x across all time points rather than at time t?

(Lines 362-363) The authors mention that the hidden states can be estimated using q distribution - which estimates/summary statistics are used here?

(Lines 367-368) What is the interpretation of the original latent states without downstream dimension reduction and clustering procedures, or they just have no interpretations?

(Line 439) This sentence is incomplete after “linear state-space model”.

(Lines 443-445) What do the authors mean by “the 8-dimensional latent states”? Do they correspond to the 8 drugs?

(Table 3) Please provide more details regarding what MinMax method and Zero-order spline method are and how you applied them in this study. Also, by saying max. and min. blood pressure, do they mean systolic and diastolic blood pressure?

Reviewer #5: Dr. Suad Ghaben

PONE-D-25-02277

A New Deep state-space Analysis Framework for Patient Latent State Estimation and Classification from EHR Time-Series Data.

Dear Authors

Please find my comments on your manuscript below. Your manuscript is an observational study that analyzed patients’ data and identified a framework for predicting the latent state of chronic disease based on the HER time-series data. Your results highlighted the potentially important temporal risk factors related to patients with cancer undergoing chemotherapy. The identified framework would help to deepen the understanding of disease progression and supports early treatment adjustments, prognostic evaluations, and the formulation of optimal long-term strategies. I commend the authors on completing this essential work.

I will start by commenting on the manuscript parts in order highlighting major and minor revisions.

Title:

Short, concise, but not reflective enough. Please add patients with cancer to the title, and no need to include the “New”.

Suggested title: The Deep state-space analysis framework for estimating the Latent state for patients with cancer undergoing chemotherapy by utilizing the HER time-series data.

Abstract:

• PLOS ONE use structured abstract. Add the headings: Introduction, Materials and methods, Results, Discussion, and amend accordingly.

• Line 23: use “identified” instead of “proposed”, as you already identified and validated the framework using the HER time-series data of 12,695 patients.

Introduction:

• Line 74: In this study, we applied this framework to the EHR of 12,695 patients. Which framework? be specific and identify the framework.

• Lines 78-81: advantages of the identified framework can be mentioned in the results, discussion and conclusion sections, rather than introduction. In the introduction, you focus on the need and rational for developing the framework. You may paraphrase this section and highlight the advantages of the identified framework in the results, discussion and conclusion sections.;

Methods:

• Lines 267- 447: Move the methods section ahead to the Results section and keep consistency of their subheadings.

• Lines 340- 364: it’s unclear whether the described “Deep state-space model” if from literature or the new model that was applied in this study. Cite where required to highlight literature, and paraphrase to highlight your work in this study.

• I suggest add a paragraph to highlight the former deep stat-space model and the new one. You may depict a figure\diagram for more clarity.,

Results:

• Written well. No comments.

• keep consistency of the methods and results sections.

Discussion:

• written well. The interpretation is rational.

Conclusion:

• written well. No comments.

General comments:

• The manuscript is lengthy; please revise for writing in a more concise language.

• The citation style is not aligned with PLOS ONE guidelines; change the citation style to Vancouver.

• The manuscript should be thoroughly revised and rearranged; move the methods section ahead to the results section and keep consistency of subheadings for both sections.

Reviewer #6: In this paper, Authors developed a new framework called the "Deep state-space analysis framework" to enable the clinical interpretation of patients and the identification of temporal risk factors in the latent state space by explicitly modeling the temporal changes of patient latent states. Authors used endpoints of latent state space as indicators for cancer patients’ prognosis outcomes. They estimated state transition probabilities and identified key risk factors that are distinguishable among the three states. They also compared the new framework to several other latent space embedding methods.

Overall, the manuscript is thorough, but some aspects of the statements lack quantitative analyses. Here are some of my comments and questions:

1. The proposed framework lacks proof of quantitative validation, i.e., what is the model’s performance in making predictions?

2. Based on the metadata, what is the state transition difference between different cancer types, gender, age group, etc.?

3. The comparison to other latent space embedding methods lacks quantitative results. UMAP is not a plausible means of comparison for clustering.

4. The states transition plots, e.g. Fig 2c and 4b, are difficult to follow, thus require clarification and better interpretation.

5. The bubble plot, e.g. Fig 3a lacks figure legends. What does the difference between different sizes of the dots imply? On what scale?

6. Can you find any results in the literature that supports your conclusions in the identified key risk factors, or vice versa?

**********

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

Reviewer #2: No

Reviewer #3: No

Reviewer #4: No

Reviewer #5: Yes: Suad Ghaben

Reviewer #6: 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.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/ . PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org . Please note that Supporting Information files do not need this step.

Dear Dr. Okuno,

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 Sep 19 2025 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.

• A rebuttal 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 .

We look forward to receiving your revised manuscript.

Kind regards,

Zeheng Wang

Academic Editor

PLOS ONE

Journal Requirements:

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.

Additional Editor Comments:

Unfortunately, several reviewers still suggest Major Revision - usually a 2nd round of Major means Rejection. However, I can see the issues raised are potentially addressable by another round of revision. I therefore suggest that the authors revise their MS again by taking all comments seriously, otherwise it is hard to guarantee an Acceptance, if, after an unsatisfactory revision.

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

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

Reviewer #1: All comments have been addressed

Reviewer #2: (No Response)

Reviewer #3: (No Response)

**********

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

Reviewer #1: Yes

Reviewer #2: Partly

Reviewer #3: Partly

**********

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

Reviewer #1: I Don't Know

Reviewer #2: No

Reviewer #3: Yes

**********

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

The PLOS Data policy

Reviewer #1: Yes

Reviewer #2: (No Response)

Reviewer #3: No

**********

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

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Yes

**********

Reviewer #1: Generally I think my previous comments were addressed in the new draft. The authors have revised and responded well to the related comments. Ok to proceed.

Reviewer #2: 1. Cluster I contains ~39% of surviving patients, and Cluster II contains ~36% of deceased patients. The manuscript defines Cluster II as an intermediate state, but it is unclear why Cluster I is not also considered intermediate?

2. The reliability of the proposed model needs to be addressed.

3. The parameters presented in tables and figures need clearer definitions and explanations, either in clinical or theoretical terms. Readers unfamiliar with the domain may not fully understand their significance without further context.

4. The manuscript would benefit from clearer writing, particularly in explaining the clinical implications and the interpretation of both data and model output.

Reviewer #3: While the authors have improved their manuscript and figures. There are remaining points to resolve:

• I still do not see any sample data in the GitHub repository. I do see mention of a sample.csv file, but there isn’t simulated or sample data to utilize with the code. Since this manuscript is largely based on a novel computational method. The methods need to be reproducible, and I cannot recommend for publication without this resolved.

• I am unsure of the contributions and impact of this work to existing deep-learning methods. I understand the utility of estimating latent states from time series data, however, findings stemming from these methods lack explainability and without comparison with another dataset, it is difficult to know if there are cohort/dataset level factors driving the identification of the latent states.

• I recommend that the authors consult with clinicians to highlight ways in which this method can improve the existing clinical paradigm of cancer treatment. For example, could the Markov transition probabilities be used in a clinical algorithm? How sure could they be that the latent state identified is probable?

**********

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

Reviewer #2: No

Reviewer #3: 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.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/ . PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org . Please note that Supporting Information files do not need this step.

Dear Dr. Okuno,

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 Nov 29 2025 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.

• A rebuttal 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 .

We look forward to receiving your revised manuscript.

Kind regards,

Zeheng Wang

Academic Editor

PLOS ONE

Journal Requirements:

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.

Additional Editor Comments:

The reviewer has pointed out the key shortcomings that should be addressed. Kindly revise your manuscript accordingly. Please note that this is the last chance to revise your manuscript, as it's a 2nd Major revision. Any improper revisions will lead to a Rejection.

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

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

Reviewer #2: (No Response)

**********

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

Reviewer #2: Partly

**********

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

Reviewer #2: No

**********

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

The PLOS Data policy

Reviewer #2: (No Response)

**********

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

Reviewer #2: Yes

**********

Reviewer #2: Thank you to the authors for their efforts in revising the manuscript and addressing the previous comments.

The authors appear to attempt to include both clinical and technical aspects in the manuscript, but the emphasis is clearly more on the technical side. Therefore, the clinical information regarding drug usage, control conditions, and medical tests should be briefly but clearly described. It would be important to explain why only the drugs or factors/medical tests mentioned in the text were included. As it currently stands, readers must accept these choices at face value without sufficient supporting justification. This issue is particularly relevant because many potential factors could be considered in such analyses. While omitting certain medical tests may improve model performance, such omissions may not be clinically appropriate or justifiable.

The comparison with other approaches should also specify how those methods were framed and implemented, not merely their final results.

An existing time-series model was used, trained, and tested with current data. However, this part is not clearly described in the manuscript. If the model was not retrained, it should be explicitly stated that testing was performed directly on existing data. In that case, an analysis of the model’s applicability and validity under these conditions is needed.

A high proportion of missing values (>50%) may cause instability in latent state estimation, particularly during the early stages of the time series. It should be analyzed whether the imputation procedure introduced bias, especially for laboratory measurements with large temporal fluctuations. The initial latent state estimation heavily depends on the imputed values; therefore, substantial imputation bias can affect model convergence and degrade the quality of the estimated latent states.

The stability and convergence of the model training process should be demonstrated, particularly given the high proportion of missing values. It would also be helpful to include additional quantitative metrics to assess model stability and robustness.

The robustness validation of the UMAP results would also be helpful.

The paper analyzes all types of cancer; how are the results for specific single cancer subtypes?

**********

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 #2: 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.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/ . PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org . Please note that Supporting Information files do not need this step.

Dear Dr. Okuno,

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 Feb 08 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.

• 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 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 .

We look forward to receiving your revised manuscript.

Kind regards,

Zeheng Wang

Academic Editor

PLOS One

Journal Requirements:

1. 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.

2. 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.

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

Reviewer's Responses to Questions

Comments to the Author

Reviewer #2: (No Response)

**********

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

Reviewer #2: Yes

**********

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

Reviewer #2: Yes

**********

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

The PLOS Data policy

Reviewer #2: (No Response)

**********

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

Reviewer #2: Yes

**********

Reviewer #2: We thank the authors for their revisions.

The reported transitions between clusters for deceased and surviving patients (lines 359–361) appear inconsistent with the figure, and the authors should clarify this discrepancy.

The robustness analysis using SDE-based synthetic data and ARI/NMI metrics is useful, but it should be noted that it only covers MCAR scenarios and may not fully reflect real-world EHR missingness patterns.

**********

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 #2: 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.

A Deep state-space Analysis Framework for Cancer Patient Latent State Estimation and Classification from EHR Time-Series Data

PONE-D-25-02277R4

Dear Dr. Okuno,

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,

Zeheng Wang

Academic Editor

PLOS One

Additional Editor Comments (optional):

Reviewers' comments: