Dear authors,

thanks a lot for submitting your revised version of your ms. It is really good and I'll be delighted to recommend it soon. In the meantime, I pass you a commented/corrected version of the pdf so that you can make the paper "perfect" for the recommended version (you should have ~135 comments in the pdf; if not, please email me so that I can pass it on to you in another way).

On top of the comments in the pdf, here are some thoughts I jotted while reading:

1. A comment:
"However, it is unlikely any statistical approach can ever solve the problem of an unmeasured explanatory variable." (page 33, in the discussion)
This is the purpose of some latent variable or latent factor models, but it's true it is not always very well done. The literature on Structural Equation Models could be enlightening regarding this aspect. For a very general introduction to plausible causality inferred from observational data, I'd suggest reading the book of Pearl:
Pearl, J. (1988) Probabilistic Reasoning in Intelligent Systems: Networks of Plausible Inference, Morgan Kaufmann.

2. General suggestions on typos, formatting, etc.
* check some statistical terminology (e.g. "marginal means" in place of "contrasts"), I have pointed out some places in the text where it sounded weird, but you might also check for other terms (it never hurts);
* words in figures look misplaced sometimes -- check that the change from doc to pdf does not add some jittering of letter location;
* formatting of refences: remove double brackets, and if necessary insert square brackets inside normal ones (but not for references, use commas instead) -- this is easy to deal with this issue in some softwares (like endnote), but there's always the last resort possibility of transforming "code" (such as citation insertions) into pure text to format it as you wish;
* also regarding formatting references: no need to call for first names (or abbreviations of first names) when citing papers;
* (language) beware of the among/across difference. Among = between entities that are part of an identified group; across = throughout a whole range of (heterogeneous) entities. You could write "among host species" or "across the range of host species", but probably not "across host species". I've pointed some places where this is misused, but I certainly have missed issues like these.

Thanks again for all the work you have done.

Sincerely,

François Massol

Dear authors,

thanks a lot for the thorough revisions you completed. As you will read, the two reviewers (one from last round of reviews, the other new) much appreciated the current version. I agree with them and I think that after some minor revisions (i.e. considering the points raised by R. Pigeault), your paper will be ready to be recommended.

I personally thank you for comparing the drop1 ANOVA tests with the model comparison procedure. However, from the results you show, I would really recommend using model comparisons (e.g. replacing drop1 tables by AICc tables of models or weight tables of variable importances) rather than drop1. From a purely statistical viewpoint, the comparisons that drop1 is making are not very relevant because they are grounded in the assumption that almost all tested variables should be included in the model. This can lead to spurious estimates of model coefficients, even if the "best model" and the drop1 tables seem to agree.

I hope I will soon read the revision of this excellent paper.

Sincerely,

François Massol

******

Additional comments from the Managing board 

Mandatory modifications 
1- Please make sure that: 
-Data are available to readers, either in the text or through an open data repository such as Zenodo (free), Dryad or some other institutional repository. Data must be reusable, thus metadata or accompanying text must carefully describe the data.   -Details on quantitative analyses (e.g., data treatment and statistical scripts in R, bioinformatic pipeline scripts, etc.) and details concerning simulations (scripts, codes) are available to readers in the text, as appendices, or through an open data repository, such as Zenodo, Dryad or some other institutional repository. The scripts or codes must be carefully described so that they can be reused.   -Details on experimental procedures are available to readers in the text or as appendices. Include information about ethical approval for animal experimentation. Provide information about the compliance of their work with ethical standards of their national ethical committees and report the reference number of the ethical committee approval. If the study did not require ethical approval, include some sentences explaining why the approval was not needed. 
-Authors have no financial conflict of interest relating to the article. The article must contain a "Conflict of interest disclosure" paragraph before the reference section containing this sentence: "The authors of this preprint declare that they have no financial conflict of interest with the content of this article." If appropriate, this disclosure may be completed by a sentence indicating that some of the authors are PCI recommenders: "XXX is one of the PCI Ecology recommenders." 
2- Please make the following changes: 
-Add the following sentence in the acknowledgements: "Version 3 of this preprint has been peer-reviewed and recommended by Peer Community In Ecology (https://doi.org/10.24072/pci.ecology.100071)" 
-If you use bioRxiv to post your preprint, add this latter sentence also in the “revision summary” section of the deposit form of bioRxiv. 
Note that this DOI is not the DOI of your article, but the DOI of the recommendation text. The DOI of your article remains unchanged. 
3- If not yet done, please send us a picture for which you own the rights that could serve as a thumbnail/illustration for your article on the web site of PCI. It can be a figure of the article.

Optional instructions (we strongly advise you to follow them) 
1- We suggest you to remove line numbering from the preprint and put the tables and figures within the text rather than at the end of your MS. 
2- Then, we strongly advise you to use the PCI templates (word docx template or latex template) to format your preprint in a PCI style. Here is the links of the templates: https://peercommunityin.org/templates/ 
→ For word template:  
Do not hesitate to modify the template as you want (and send it back to us if you made significant improvements). 
-the text to be replaced by your own text starts with XXX, eg XXXXTitle of the article. 
-XXXXthe "citeas"  → Edeline, E. and Loeuille, N. (2021) Size-dependent eco-evolutionary feedbacks in fisheries. bioRxiv, 2020.04.03.022905, ver. 3 peer-reviewed and recommended by PCI Ecology. doi: https://doi.org/10.1101/2020.04.03.022905 
-XXXXthe date of deposit in the preprint server  →  date of the deposit of the latest version 
-XXXXthe surnames and names of the reviewers we sent you  → Jean-François Arnoldi and an anonymous reviewer 
-XXXXthedoiwesentyou →  https://doi.org/10.24072/pci.ecology.100071 
-XXXXthe surname and name of the recommender → Simon Blanchet 
-In the acknowledgements, add this sentence → "Version 3 of this preprint has been peer-reviewed and recommended by Peer Community In Ecology (https://doi.org/10.24072/pci.ecology.100071)" 
-Please be careful to choose the badges “Open Code” and “Open Data” only if appropriate (in addition to the “Open Access” and “Open Peer-Review” badges).  
→ For Latex and mode org templates: 
Do not hesitate to modify the template as you want (and send it back to us if you made significant improvements). 
-main.tex and sample.bib should be filled.  
-in main.tex, the recommender’s name is "Simon Blanchet" and the reviewers’ names are Jean-François Arnoldi and an anonymous reviewer -In sample.bib, indicate the right version of your preprint. It is version 3 
-Preambuleecology.tex should be modified (comment lines 115, 119) to select badges. Please be careful to choose the badges “Open Code” and “Open Data” only if appropriate (in addition to the “Open Access” and “Open Peer-Review” badges).  
3- we suggest that you deposit a copy of your MS in zenodo.org and ask for its inclusion in the PCI community (“Communities” section in the deposit form). Indicate the current doi of your MS, if it already has one, in the “doi” section.

Dear authors,

with two reviews and my own reading of your paper, I would like you to conduct a thorough revision before I can recommend your work. As you will see, both reviewers found a lot of merits to your work, but they also raised some concerns, notably regarding the limitations of the methods and their interpretation, as well as the lack of justification for some methodological choices.

Personally, I also find merits to your manuscript: the questions and data are sound, the general approach is interesting and the results are going to be quite useful for applied ecologists working on zoonotic prevention. However, you I have some issues with some of your methodological choices:

1. A simple one: PCI asks for data and code availability, but I have not been able to find any (or maybe my eyes don't work that well these days). Can you make R scripts and the raw data available for the next version of your manuscript?

2. For virus and bacterial detection, you use two different procedures (antibody detection and 16S amplification and sequencing). However, these are not only technical differences, but also epistemic ones: with antibodies, you are only going to find viruses that you were looking for, while with 16S you can find bacteria that you even had not suspected in the first place. For this reason, many of the analyses will not behave in the same way for viruses and bacteria. I suggest that you acknowledge this early on (in the M&M) and discuss it thoroughly.

3. While I understand your focus on pathogenic bacteria and viruses, it would probably be interesting to also look at other forms of symbiont association (negative or positive), for instance non-pathogenic bacteria with pathogenic ones, non-pathogenic viruses with pathogenic bacteria, etc. You obviously did not search for non-pathogenic viruses like bacteriophages, but at least if you still have blasted data on non-pathogenic bacteria, there might be something to dig in there about microbial associations within rodents (notably, it would be cool to know whether some other bacteria are negatively associated with nasty pathogens).

4. Quite a large part of the Introduction announces apicomplexan, protozoa, cestods, trematods, etc. as important parasites of rodents, but you chose to focus on bacteria and viruses. Why? (I guess for money reason, but maybe you can elaborate)

5. For the diversity analyses, did you use multiplicative or additive Shannon numbers? (i.e. the exponential version or the log version) As you considered both beta and alpha diversities, which version of the partitioning did you use? (as there are at least two main ones, Jost-Routledge's and Chao's) Did you correct for the usual bias in estimates of Shannon diversity? (using e.g. Chao's estimator) I think the entropart package does this...

6. I'm not a big fan of the drop1 method (i.e. likelihood ratio test with the complete model only) because the best model explaining your data can actually have fewer than n-1 variables in it (if n is the total of all tested variables). A very simple trick would be to run your complete model through GLMulti to get the AIC or BIC of all models and ascertain which one is the best. That way, you could have a plausible answer to the question "which factors affect Shannon diversity of pathogens?".

7. At some point you mention that you wanted to relate beta diversities with factors, but I did not see that exactly. An option to do this easily is to use dissimilarities obtained from beta diversities (see e.g. Ohlmann et al. 2019) as a distance and then make an adonis analysis on these distances using factors as covariables (adonis is in package vegan). That way, you could explain differences in pathogen composition using some of your factors (habitats...).

8. Fig. 2 has an issue: I don't think you can test anything using residuals. If you delete one factor in a model and then plot the residuals according to the modalities of that dismissed factor, what you will see is only how bad your model is at compensating this missing factor, not the intrinsic differences between modalities. To obtain the differences between modalities, you need to actually predict the model (with the factor you want to look at) on new data (using predict in R), probably with a randomization (bootstrap) of the initial data.

9. In Fig. 3B and around it: when you write "pathogen co-exposure", does it mean "virus + bacteria co-infections"?

10. Regarding the "MCA -> principal dimensions to be regressed individually by factors" approach, I'm sure you can do better than that by using CCA (Peres-Neto et al. 2006) or distance-based RDA (Blanchet et al. 2014) to obtain the fractions of the total chi² of your host x pathogen table explained by each factor. The individual regression of MCA eigenvector seems a bit pedestrian and it does not give a super-clear answer in the end.

11. Fig. 4 does not seem that useful -- do you really need it?

12. As stated above about model comparisons through AIC or BIC using GLMulti, I suggest you do something like that for the GLM you do on pathos-pathos associations.

Or maybe better: for this part, you can also shift completely and use your dataset as a network (host nodes connected to pathos nodes). By using any sensible community-detection algorithm or block model, you can find modules/blocks/communities of pathogens that have similar connection patterns, which in a way is what you are looking for, isn't it? This approach is less driven by hypotheses (it's actually super-exploratory), but it might confirm what you know while at the same time finding associations you did not see or did not expect. I suggest Leger et al. (2015) as a good reading on modules/blocks in networks, if needed.

Or you can do both: GLMs to test particular associations and module-search to get a broader view on all associations...

1. A very general comment: you mention p-value corrections only for Tukey's tests, but given the amount of tests performed, I suggest you use p.adjust with Benjamini-Hochberg correction overall (so you do that for the vector of all p-values obtained with all analyses on the same dataset).

2. The discussion is a bit long (9 pages +), I guess you can shorten the parts where you repeat the results.

Typos/minor things L. 90 Leptospirosa => Leptospira ? L. 133 sexually-immature => sexually immature L. 208 "This produced a down to a set of quantitative (...)" The sentence looks warped. L. 252 implicated => involved ? L. 276-277 For M&M instead? L. 290-297 Problems with punctuation and coherence between propositions: "because they were...", "inability to rule out...", "or because their identity..." L. 335 pseudo-R² from deviance ? Please state it. L. 424 "more rarely than expected (p = 0.13..." -- really expected? A p-value above 0.05 more likely means "as expected"

I hope you will find all reviews useful for revising your paper. Sincerely,

François Massol

Cited references Blanchet, F. G., Legendre, P., Bergeron, J. A. C. & He, F. (2014) Consensus RDA across dissimilarity coefficients for canonical ordination of community composition data. Ecological Monographs, 84, 491-511.

Chao, A. & Chiu, C.-H. (2016) Bridging the variance and diversity decomposition approaches to beta diversity via similarity and differentiation measures. Methods in Ecology and Evolution, 7, 919-928.

Chao, A., Wang, Y. T. & Jost, L. (2013) Entropy and the species accumulation curve: a novel entropy estimator via discovery rates of new species. Methods in Ecology and Evolution, 4, 1091-1100.

Jost, L. (2007) Partitioning diversity into independent alpha and beta components. Ecology, 88, 2427-2439.

Leger, J.-B., Daudin, J.-J. & Vacher, C. (2015) Clustering methods differ in their ability to detect patterns in ecological networks. Methods in Ecology and Evolution, 6, 474-481.

Ohlmann, M., Miele, V., Dray, S., Chalmandrier, L., O’Connor, L. & Thuiller, W. (2019) Diversity indices for ecological networks: a unifying framework using Hill numbers. Ecology Letters, 22, 737-747.

Peres-Neto, P. R., Legendre, P., Dray, S. & Borcard, D. (2006) Variation partitioning of species data matrices: Estimation and comparison of fractions. Ecology, 87, 2614-2625.

Routledge, R. D. (1979) Diversity indices: Which ones are admissible? Journal of Theoretical Biology, 76, 503-515.