Dear Dr. Kim,

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 Apr 03 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.. 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.. 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.. 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 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 . 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 . 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 . 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,

Zixiang Gao

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.

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.

[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: Partly

Reviewer #2: Partly

**********

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

Reviewer #1: Yes

Reviewer #2: 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.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

Reviewer #2: No

**********

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

Reviewer #1: Yes

Reviewer #2: No

**********

Reviewer #1: This manuscript proposes a multi-task gradient boosting framework (MTGBM) that integrates multi-modal molecular representations to jointly predict human clearance (CL) and volume of distribution (Vd). The topic is timely and relevant, as accurate early-stage PK prediction remains a major bottleneck in drug discovery. The authors combine CNN-derived molecular image embeddings, descriptor- and fingerprint-based MLP embeddings, BorutaSHAP feature selection, and a custom multi-output gradient boosting objective. While the study is carefully executed and technically sound, its impact is substantially limited by presentation issues, modest empirical gains, and an overstatement of novelty relative to existing machine-learning literature. Overall, the work may be suitable for a general-scope journal, but it requires significant revision in presentation, positioning, and contextual discussion to meet current standards.

The presentation quality is a serious weakness. Many figures use extremely small fonts that are difficult to read even when zoomed in, particularly the BorutaSHAP plots and scatter plots comparing single-task and multi-task models. Given that interpretability (via SHAP) is emphasized as a key contribution, the unreadable axis labels, legends, and annotations significantly undermine the significance of the findings. This is not a cosmetic issue: poor readability directly affects the readers' ability to assess the claims. All figures should be regenerated with substantially larger fonts, simplified legends, and clearer layouts. In addition, some figures appear redundant (e.g., multiple scatter plots across splits conveying similar information) and could be consolidated to improve clarity.

The claimed novelty of the work is overstated. Multi-task gradient boosting machines with shared tree structures are not new and have already been formally introduced and analyzed. Moreover, from a practical standpoint, multi-output or multi-task tree-based models are well supported in widely used ecosystems such as LightGBM and scikit-learn, either directly or through standard wrappers. While it is true that such approaches are less commonly applied in PK prediction, this constitutes an application-level novelty rather than a methodological one.

Similarly, the construction of 'multi-task gradient boosting trees' in this manuscript largely follows established paradigms and does not introduce fundamentally new optimization strategies, regularization schemes, or task-coupling mechanisms. The improvements reported are modest (4.9% MSE reduction for CL and 10.8% for Vd) and, importantly, not statistically significant under the Diebold–Mariano test. As such, the authors should significantly temper their novelty claims and reframe the contribution as a careful application and validation study rather than a methodological advance.

A major omission is the lack of discussion and comparison with recent blind-prediction challenges and benchmarks that strongly support the superiority of multi-modal, multi-task pretraining regimes. For example, recent work reported in Journal of Chemical Information and Modeling 65 (19), 10465-10476 demonstrates that large-scale, multi-modal, multi-task pretraining consistently outperforms task-specific models on ADME-relevant endpoints, especially in low-data regimes. Conceptually, the current MTGBM framework addresses a similar motivation but does so in a comparatively shallow, non-pretrained setting.

The authors should explicitly discuss how their approach relates to, differs from, and potentially complements such pretraining-based frameworks. In particular, it would be valuable to clarify whether the modest gains observed here stem from limited dataset size (~700 compounds) and whether the approach would scale or remain competitive in settings where pretrained representations are available.

Relatedly, the manuscript does not engage with the rapidly growing body of work on large models and foundation models for molecular and drug discovery. Recent studies (e.g., arXiv:2511.11257) have demonstrated that large, pretrained models (often trained across millions of molecules and multiple objectives) can achieve outstanding predictive power on downstream PK and ADME tasks, frequently surpassing classical ML pipelines even without extensive feature engineering.

Given this context, the authors' reliance on relatively small CNN and MLP embeddings trained specifically for CL and Vd should be critically discussed. Are the gains from MTGBM expected to persist when stronger, pretrained representations are used? Could the proposed multi-task boosting layer act as a lightweight downstream learner on top of foundation-model embeddings? Without such discussion, the work risks appearing disconnected from the current trajectory of the field.

Reviewer #2: This study presents a novel approach to predicting key pharmacokinetic parameters by combining multi-modal molecular representations with multi-task learning. The authors have clearly put significant thought into designing their MTGBM framework, and the integration of different feature types including CNN-based structural embeddings, molecular fingerprints, and physicochemical descriptors represents a thoughtful attempt to capture complementary information about drug molecules. The use of BorutaSHAP for feature selection is also a nice touch that adds rigor to the modeling process.

However, I do have some concerns about the work that should be addressed before publication. The most significant issue is the lack of statistical significance in the improvements over single-task models. While the authors acknowledge this limitation, they somewhat overstate the potential of their approach throughout the manuscript. The Diebold-Mariano test results with p-values of 0.117 and 0.245 for CL and VD respectively indicate that we cannot rule out the possibility that the observed improvements occurred by chance. Given that the dataset contains only about 700 compounds, this is perhaps not surprising, but the discussion should more carefully frame these findings as preliminary evidence rather than definitive proof of the method's superiority.

The dataset size itself poses another challenge. With only around 700 compounds split into training, validation, and test sets, the test set ends up being quite small at 139 samples. This raises questions about the stability of the performance metrics and whether they might be heavily influenced by a few outliers. The authors mention using a robust optimization algorithm for splitting, but it would be helpful to know more about how stable the results were across different random splits.

I also have some methodological concerns about the feature extraction process. The authors trained separate CNN and MLP models for CL and VD predictions before extracting embeddings for the MTGBM model. This creates a potential information leakage issue since these embedding models were trained using the same data that later gets used for training and evaluating the MTGBM. Ideally, the embedding extraction should be done in a way that keeps test data completely separate, perhaps through nested cross-validation or by training the embedding models only on training data before any validation or testing occurs.

The hyperparameter optimization approach for the MTGBM model is another area that needs clarification. The authors mention using "estimated optimal parameters" from single-task models due to computational constraints. While this is understandable, it does mean we cannot be sure that the MTGBM model is performing at its full potential. A properly optimized multi-task model might show larger improvements, or conversely, the current results might be underestimating its true performance. Either way, this limitation should be more prominently discussed.

The interpretability analysis using SHAP values raises some questions about what exactly the MLP and CNN embeddings represent. These are learned representations from neural networks trained specifically to predict CL and VD, so it is somewhat circular when they emerge as important features in the SHAP analysis. The authors interpret this as evidence of successful cross-task information sharing, but it might simply reflect that these embeddings already contain target-specific information from their training process. A cleaner approach might have been to use unsupervised molecular embeddings that were not trained on the prediction tasks themselves.

I noticed some inconsistencies in the reported results that need attention. For VD prediction, the MSE improved substantially from 0.950 to 0.848, yet the GMFE increased from 1.844 to 1.938. The authors suggest this indicates the importance of evaluating multiple metrics, which is valid, but this pattern deserves more discussion. Typically, improvements in MSE would be expected to correspond with better GMFE, so the divergence here might indicate something interesting about the error distribution that could be explored further.

The manuscript could also benefit from more careful editing in places. The literature review section repeats some content between sections 2.1 and 2.2, and there are minor formatting inconsistencies in the references. The figures mentioned in the text, particularly Figures 1 through 6, are referenced but not actually included in the manuscript file, making it impossible to fully evaluate the visual presentations of the results.

Despite these concerns, the work makes a valuable contribution by demonstrating how multi-task learning can be applied to pharmacokinetic prediction. The idea of leveraging correlations between related parameters is conceptually sound and worth exploring further. The authors have been transparent about their limitations, which is commendable. With some revisions to address the methodological issues and a more nuanced interpretation of the results, this could become a useful addition to the literature on computational pharmacokinetics. The framework they have developed could potentially be extended to other ADME properties, and their careful documentation of the modeling process provides a good foundation for others building on this work.

**********

what does this mean?). If published, this will include your full peer review and any attached files.). If published, this will include your full peer review and any attached files.). If published, this will include your full peer review and any attached files.). 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 For information about this choice, including consent withdrawal, please see our For information about this choice, including consent withdrawal, please see our For information about this choice, including consent withdrawal, please see our Privacy Policy..-->

Reviewer #1: Yes: Zhaoxi SunZhaoxi SunZhaoxi SunZhaoxi Sun

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.

Multi-Task Gradient Boosting with Multi-Modal Molecular Representations for Simultaneous Prediction of Drug Clearance and Volume of Distribution

PONE-D-25-68444R1

Dear Dr. Kim,

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  and clicking the ‘Update My Information' link at the top of the page. For questions related to billing, please contact  and clicking the ‘Update My Information' link at the top of the page. For questions related to billing, please contact  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,

Zixiang Gao

Academic Editor

PLOS One

Additional Editor Comments (optional):

Thank you for your patience and for carefully addressing the reviewers' comments. We are pleased to accept your manuscript for publication in PLOS ONE.

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

Reviewer #1: All comments have been addressed

Reviewer #2: All comments have been addressed

**********

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

Reviewer #1: Partly

Reviewer #2: Yes

**********

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

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. 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.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

Reviewer #2: (No Response)

**********

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

Reviewer #1: Yes

Reviewer #2: Yes

**********

Reviewer #1: (No Response)

Reviewer #2: Accept

Accept

Accept

Accept

Accept

Accept

AcceptAcceptAcceptAcceptAcceptAcceptAcceptAcceptAcceptAccept

**********

what does this mean?). If published, this will include your full peer review and any attached files.). If published, this will include your full peer review and any attached files.). If published, this will include your full peer review and any attached files.). 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 For information about this choice, including consent withdrawal, please see our For information about this choice, including consent withdrawal, please see our For information about this choice, including consent withdrawal, please see our Privacy Policy..-->

Reviewer #1: Yes: Zhaoxi SunZhaoxi SunZhaoxi SunZhaoxi Sun

Reviewer #2: No

**********