Dear Dr. Youngblut,

Thank you very much for submitting your manuscript "Interpreting tree ensemble machine learning models with endoR" for consideration at PLOS Computational Biology.

As with all papers reviewed by the journal, your manuscript was reviewed by members of the editorial board and by several independent reviewers. In light of the reviews (below this email), we would like to invite the resubmission of a significantly-revised version that takes into account the reviewers' comments.

The reviewers generally consider that the work is a good contribution to the field and the manuscript is well-written. However, they also raise some points that should be addressed. In particular, the concern of reviewer #1 as to how the feature selection is performed is most worrisome as it is an issue of correctness.

We cannot make any decision about publication until we have seen the revised manuscript and your response to the reviewers' comments. Your revised manuscript is also likely to be sent to reviewers for further evaluation.

When you are ready to resubmit, please upload the following:

[1] A letter containing a detailed list of your responses to the review comments and a description of the changes you have made in the manuscript. Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.

[2] Two versions of the revised manuscript: one with either highlights or tracked changes denoting where the text has been changed; the other a clean version (uploaded as the manuscript file).

Important additional instructions are given below your reviewer comments.

Please prepare and submit your revised manuscript within 60 days. If you anticipate any delay, please let us know the expected resubmission date by replying to this email. Please note that revised manuscripts received after the 60-day due date may require evaluation and peer review similar to newly submitted manuscripts.

Thank you again for your submission. We hope that our editorial process has been constructive so far, and we welcome your feedback at any time. Please don't hesitate to contact us if you have any questions or comments.

Sincerely,

Luis Pedro Coelho

Associate Editor

PLOS Computational Biology

Dina Schneidman

Software Editor

PLOS Computational Biology

***********************

The reviewers generally consider that the work is a good contribution to the field and the manuscript is well-written. However, they also raise some points that should be addressed. In particular, the concern of reviewer #1 as to how the feature selection is performed is most worrisome as it is an issue of correctness.

Reviewer's Responses to Questions

Comments to the Authors:

Please note here if the review is uploaded as an attachment.

Reviewer #1: Ruaud and colleagues present endoR, an R package aimed to enable better interpretation of tree ensemble methods, particularly for microbiome data. They showcase the utility of their package using fully simulated data, artificial phenotypes (real data with a simulated response variable), and two real-world applications (prediction of liver cirrhosis from fecal metagenomes and prediction of the presence of Methanobacteriaceae in the gut). The software addressed a well-known problem in the field (i.e. limited interpretability of complex tree-ensemble machine learning methods) and their article is well written (albeit a bit verbose in places), but several questions remains, especially regarding feature selection, comparisons to other tools/measures and comparisons to other machine learning methods.

Feature selection: It sounds like the feature selection step was performed before model training (I hope I did not miss a detailed description of the feature selection, but in any case I would advise the authors to describe their feature selection procedure more prominently and in more detail). This can lead to overfitting and over-optimistic models and should be avoided (see for example https://www.ncbi.nlm.nih.gov/pubmed/19880370).

Comparison to other methods: The comparison to other methods (Gini coefficient, SHAP values, etc.) is the part of the manuscript that could be improved the most. Figure 2G indicates that endoR might perform better than the Gini coefficient, but it is only shown for the top features in the main text. Similarly, Figure 4 focuses on the top 5 false positive features only (disregarding the reconstructed network for now). I think it would be possible to construct AUROC measures for how well Gini coefficients, SHAP values, and endoR can recover the true underlying features (in repeated(?, see below) AP simulations), covering all features by evaluating the true positive and false positive rates, even though closely related features would have to be disregarded or evaluated separately.

Comparison to other machine learning methods: There are also other machine learning algorithms except the random forest routinely used in the microbiome field. For example, both LASSO or Elastic Net can have similar (or even slightly better) performance to the random forest (https://pubmed.ncbi.nlm.nih.gov/33785070/, given properly normalised input data), and both of these do not suffer of limited interpretability. The authors should compare their endoR importance measures also against features weights derived from those more white-box models.

Limited interactions in cirrhosis predictions: For the cirrhosis dataset, almost no meaningful interaction between features has been detected by endoR. I think the authors should discuss this observation. Does this mean that interaction networks between features do not play a prominent role in disease predictions? Relating to the point above, would a less-complex machine learning algorithm (such as LASSO) have been sufficient? (This is probably outside the scope of this paper, but how do different diseases compare in this aspect?)

Minor questions:

The prevalence threshold of 25% seems rather high. Can the authors motivate the level of the cutoff? For example, one of the best-known predictors for colorectal cancer, Fusobacterium nucleatum, is not present at a prevalence higher than 10% in a typical case-control study, and therefore, a prevalence cutoff of 5% might be more appropriate.

The software described in this article is available on Github. Are there plans to submit the R package to the Bioconductor repository? Also, the existing vignettes are rather short and do not include many details. Can those be expanded for better usability of the software?

Was the creation of the artificial phenotype repeated several times? It seems that the fully simulated data were generated with 100 independent repetitions, but I could not find this information for the artificial phenotypes.

The references to Figure 4 on page 7 are probably referring to Figure 5 and should be adjusted

The asterisk glyphs in Figure 6E are explained in the figure legend but could also be explained in a visual figure legend (such as in Figure 6C).

Reviewer #2: In this manuscript, the authors developed an algorithm for interpreting the results from those tree-based machine learning algorithms. Tree-based ML methods have been widely used for microbiome-based prediction due to the complex characteristics of microbiome data and potential high-order interactions among the taxa. However, these ML methods often did not provide sufficient interpretation. Thus, the work is considered to be very significant. The manuscript also reads well and the experiments are carefully designed to demonstrate the method. I have some comments the authors may consider in revision.

1. The interaction importance defined as the square root of the product of individual feature importance did not seem to conform to the traditional statistical interaction, which describes the phenomenon that the effect of one explanatory variable on the outcome depends on another explanatory variable or the joint effect of two explanatory variables is different from the sum of individual effects. Given the current definition, I do not see how this is related to the traditional statistical interaction. Please elaborate.

2. How does the algorithm perform under small sample sizes? In the real world, the majority of microbiome studies have small sample sizes. Performance under small sample sizes should be studied and caveat should be given if the algorithm does not work well for small sample sizes.

3. Global null conditions should also be studied. Given that there is no relationship between the microbiome and the phenotype, what is the frequency the algorithm will return any findings, and how many false positives if making any findings?

4. In the analysis of Methanobacteriaceae phenotype, how to address the compositional effects? Since the sequencing data are compositional in nature, the increase in the abundance of one taxon will naturally lead to decreases in the relative abundances of other taxa so there is a natural negative influence on the relative abundance of Methanobacteriaceae.

5. The bootstrapping the authors describes seems to be the sub-sampling technique, not sampling with replacement.

6. It does not make much sense to compare to those network inference algorithms such as sparCC and gLasso since they are detecting different networks.

Reviewer #3: The paper by Albane Ruaud et al. presents endoR, a machine learning strategy aiming at interpreting a fitted tree ensemble model. The method is validated experimentally on both simulated and real metagenomic data. Results show that endoR is able to infer true associations with more or comparable accuracy than other state-of-the-art approaches while easing model interpretation. The software is available as an open-source R-package on GitHub (https://github.com/leylabmpi/endoR).

This is a nice contribution to the microbiome community. The manuscript is well written and structured.

I have few minor comments:

- Is the method suitable for (shotgun) metagenomic data only or could be adopted also for 16S data? Please specifiy better this aspect in the text.

- I understand that data used for the analysis are too big to be hosted in github. Could you please put them somewhere else?

**********

Have the authors made all data and (if applicable) computational code underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data and code 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 and code 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 or code —e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: None

Reviewer #3: No: Please make data used for this analysis public available

**********

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

Reviewer #2: No

Reviewer #3: No

Figure 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. 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 us at figures@plos.org.

Data Requirements:

Please note that, as a condition of publication, PLOS' data policy requires that you make available all data used to draw the conclusions outlined in your manuscript. Data must be deposited in an appropriate repository, included within the body of the manuscript, or uploaded as supporting information. This includes all numerical values that were used to generate graphs, histograms etc.. For an example in PLOS Biology see here: http://www.plosbiology.org/article/info%3Adoi%2F10.1371%2Fjournal.pbio.1001908#s5.

Reproducibility:

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. Additionally, PLOS ONE offers an option to publish peer-reviewed clinical study protocols. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols

Dear Dr. Youngblut,

Thank you very much for submitting your manuscript "Interpreting tree ensemble machine learning models with endoR" for consideration at PLOS Computational Biology. As with all papers reviewed by the journal, your manuscript was reviewed by members of the editorial board and by several independent reviewers. The reviewers appreciated the attention to an important topic. Based on the reviews, we are likely to accept this manuscript for publication, providing that you modify the manuscript according to the review recommendations.

We believe that major issues have been addressed and that the outstanding doubts of Reviewer #1 are quickly addressable.

Please prepare and submit your revised manuscript within 30 days. If you anticipate any delay, please let us know the expected resubmission date by replying to this email.

When you are ready to resubmit, please upload the following:

[1] A letter containing a detailed list of your responses to all review comments, and a description of the changes you have made in the manuscript. Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out

[2] Two versions of the revised manuscript: one with either highlights or tracked changes denoting where the text has been changed; the other a clean version (uploaded as the manuscript file).

Important additional instructions are given below your reviewer comments.

Thank you again for your submission to our journal. We hope that our editorial process has been constructive so far, and we welcome your feedback at any time. Please don't hesitate to contact us if you have any questions or comments.

Sincerely,

Luis Pedro Coelho

Academic Editor

PLOS Computational Biology

Dina Schneidman

Software Editor

PLOS Computational Biology

***********************

A link appears below if there are any accompanying review attachments. If you believe any reviews to be missing, please contact ploscompbiol@plos.org immediately:

We believe that major issues have been addressed and that the outstanding doubts of Reviewer #1 are quickly addressable.

Reviewer's Responses to Questions

Comments to the Authors:

Please note here if the review is uploaded as an attachment.

Reviewer #1: I want to thank the authors very much for addressing my previous comments. I can imagine that it was a lot of work to include the LASSO classifier in their comparison as well and I appreciate their effort. Also, I thank the authors to submitting their package to CRAN. I hope it can be useful to the wider community. Lastly, restructuring the paper improved its readability very much.

I would have some question regarding the comparison to other methods, again.

Supplementary Figure 7 and Supplementary Figure 8 EF seem to identical. Is this intended? Also, I cannot really find correspondence between Figure 4 and Supplementary Figure 8 A-D, although they should show the same data (evaluation on the AP data). I would hope that the authors can clarify my confusion here. For example, the orange dot in SFig 8A indicated that at this FPR (it should be FPR in the axis label as well), only LASSO with interactions and the Wilcoxon perform worse than endoR, but in Fig 4A, Lasso with interaction performs better than the Wilcoxon and normal Lasso performs worse (but has a similar TPR in SFig 8A). From the figure legends I would think that the same data are shown, but as rates instead of numbers. Also, I would suggest to show both rates and the numbers in the main figure. As a reader, I personally would be more interested into the TPR at a given FPR than in absolute numbers for five FP detections.

Reviewer #2: The authors have addressed my comments sufficiently.

Reviewer #3: I feel the authors have answered positively to my previous comments and modified the manuscript according to them.

**********

Have the authors made all data and (if applicable) computational code underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data and code 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 and code 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 or code —e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Yes

**********

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

Reviewer #2: Yes: Jun Chen

Reviewer #3: No

Figure 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. 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 us at figures@plos.org.

Data Requirements:

Please note that, as a condition of publication, PLOS' data policy requires that you make available all data used to draw the conclusions outlined in your manuscript. Data must be deposited in an appropriate repository, included within the body of the manuscript, or uploaded as supporting information. This includes all numerical values that were used to generate graphs, histograms etc.. For an example in PLOS Biology see here: http://www.plosbiology.org/article/info%3Adoi%2F10.1371%2Fjournal.pbio.1001908#s5.

Reproducibility:

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. Additionally, PLOS ONE offers an option to publish peer-reviewed clinical study protocols. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols

References:

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.

Dear Dr. Youngblut,

We are pleased to inform you that your manuscript 'Interpreting tree ensemble machine learning models with endoR' has been provisionally accepted for publication in PLOS Computational Biology.

Before your manuscript can be formally accepted you will need to complete some formatting changes, which you will receive in a follow up email. A member of our team will be in touch with a set of requests.

Please note that your manuscript will not be scheduled for publication until you have made the required changes, so a swift response is appreciated.

IMPORTANT: The editorial review process is now complete. PLOS will only permit corrections to spelling, formatting or significant scientific errors from this point onwards. Requests for major changes, or any which affect the scientific understanding of your work, will cause delays to the publication date of your manuscript.

Should you, your institution's press office or the journal office choose to press release your paper, you will automatically be opted out of early publication. We ask that you notify us now if you or your institution is planning to press release the article. All press must be co-ordinated with PLOS.

Thank you again for supporting Open Access publishing; we are looking forward to publishing your work in PLOS Computational Biology. 

Best regards,

Luis Pedro Coelho

Academic Editor

PLOS Computational Biology

Dina Schneidman

Software Editor

PLOS Computational Biology

***********************************************************