All reviews of published articles are made public. This includes manuscript files, peer review comments, author rebuttals and revised materials. Note: This was optional for articles submitted before 13 February 2023.
Peer reviewers are encouraged (but not required) to provide their names to the authors when submitting their peer review. If they agree to provide their name, then their personal profile page will reflect a public acknowledgment that they performed a review (even if the article is rejected). If the article is accepted, then reviewers who provided their name will be associated with the article itself.
Thank you for your quick response.
[# PeerJ Staff Note - this decision was reviewed and approved by Keith Crandall, a PeerJ Section Editor covering this Section #]
I apologize for the delay, There are some minor issues raised by reviewer 3. Could you please address them quickly and I promise I will give a quick turn around once I received your revision. Thank you.
Well written throughout, with no major errors.
The aim of the manuscript is well described, and the research gap is pointed out, i.e. that genotypes from large sequencing datasets and large phenotype datasets have not been well integrated.
The authors have addressed or at least acknowledged previous concerns about significance and power, so that they are now clear to the reader.
Again, the authors have addressed my concerns abut the validity of the findings, and pointed out the caveats in a way that is clear to the reader.
There is still some question about the novelty of the biological findings (as they are not validated), but I consider this to be secondary to the main point of the paper: the authors have produced an easily used code for looking for variants between strains that they can expand in the future, and shown that it gives useful outputs.
Dear Authors,
I feel that you have addressed the review comments well, in that you have made clear to the reader what the short-comings are, and how to minimize false discoveries.
Smaller, minor suggestions are below:
Lines 62-63: The preprint Ashbrook et al was recently published, PMID: 33472028 https://t.co/gDeKS50ANt?amp=1
Lines 69-70: I think it would read better if the sentence was “Several of the aforementioned resources allow the user to interactively query genotypes for 70 user-selected inbred mouse strains for input genes or genetic regions”.
Lines 70-72: The Variant browser in GeneNetwork has some of this functionality (http://www.genenetwork.org/snp_browser and http://gn1.genenetwork.org/snpbrowser.html), however it does not seem to be programmatic accessible, limiting its usefulness compared to MouseFM.
Line 181: Should read “we used” rather than “we use”.
Expression quantitative trait loci section: It would be interesting to know how many of the novel (i.e. not found in Mostafavi) variants are possibly cis (e.g. with ~5-10Mb of the gene)? I would expect many of the these large effect variants to be in cis. This might be another quality metric, or another way to determine the best candidate gene for expression of that gene.
Lines 263-272: Looking at the Zimmerman et al 2019 paper, it seems that they don't give confidence intervals (e.g. 1.5 LOD drop), but rather just gave the peak position. Eyeballing their QTL maps (without having the data to reanalyze it), it seems like your Chr7, and certainly your Chr11 SNPs overlap with the QTL detected in Zimmerman et al.
To me, it would be a plausible suggestion that you have identified variants underlying those two QTL. It is also possible that the two QTL that you do not see in this study are driven by the other strains that you did not use.
Lines 289-290: It seems odd that you detect the same QTL region on chr7 for Dystrophic cardiac calcification and Interfrontal bone. Is this due to chance? Biological overlap (e.g. something in the calcification/bone pathway)? Or similar strains, and therefore regions that are divergent between the groups appear often?
Line 301: “a setting” is not needed.
Line 310-311: If the data for the eQTL mapping was available, you could redo it, using only your subset of strains, to test this hypothesis.
Line 379-380: the expression datasets are also an example of quantitative traits.
Please pay particular attention to reviewer 3's comments.
[# PeerJ Staff Note: Please ensure that all review comments are addressed in a rebuttal letter and any edits or clarifications mentioned in the letter are also inserted into the revised manuscript where appropriate. It is a common mistake to address reviewer questions in the rebuttal letter but not in the revised manuscript. If a reviewer raised a question then your readers will probably have the same question so you should ensure that the manuscript can stand alone without the rebuttal letter. Directions on how to prepare a rebuttal letter can be found at: https://peerj.com/benefits/academic-rebuttal-letters/ #]
This paper presents a tool (MouseFM) that can extract genome-wide all variants that are different between two user specified groups of inbred mouse strains. They downloaded Variant data from Mouse Genomes Project and other public websites and filtered the results.
Major: The bioconductor package doesn’t work and so it could not be tested.
The author needs to provide comparison statistics about MouseFM with others .
They don’t have enough data to support their claim that their tool outperformed others. The results need to be experimental validated. No future work was mentioned
Minor:
There is no description for sub-figure 4C.
The presentation of the material is fine
The design and analysis are straight forward
Although there is some replication of previously identified candidate genes, there is no way to know if the novel candidates are any thing more that an association. This makes it difficult to tell if the novel findings are meaningful.
The approach seems to be one that is already used and not necessarily novel.
The conclusions are based on assumptions that predictions made through Vep are accurate and this may not always be the case.
The work would be strengthened if there were some independent means to demonstrate that the new fine-mapped SNP are indeed functional. This is of course a big ask - If the authors pointed out these limitations in the discussion this would certainly be an improvement
The quality of writing is generally good, with a few minor errors, and the article is formatted well. One important omission is that Panel C of figure 4 is not mentioned in the legend.
The aim of the manuscript is well described, and the research gap is pointed out, i.e. that genotypes from large sequencing datasets and large phenotype datasets have not been well integrated.
I have concerns about the methods, and how they were shown to be significant, which I have described in my general comments.
The results presented are valid, but I have concerns about how generalizable they are, as outlined in my general comments. The underlying data and code is available, and the R code is easy to use.
Dear Authors,
The manuscript is interesting, and provides a good interface for looking for variants between inbred strains. As the authors point out, there is now good phenotype and genotype data for many of these inbred strains, and these two resources should be taken advantage of. The manuscript is well written, with the authors points being clearly made.
That being said, I have a few concerns that I feel need to be addressed.
It is clear to me that this method works for mendelian traits, such as albinism used by the authors. It is much less clear how well this will work for quantitative traits, especially those that have several causal variants of similar effect size. If it is used as a second-line after QTL mapping (i.e. just looking at between strain differences within the QTL region) then I see how it would work, but as presented, it appears to be looking genome-wide from the start and clustering the strains into two groups, which would simply not work for cases with several large effect loci.
I think rather than just looking at three examples, a more convincing argument would be made if a whole dataset (e.g. all of the JAX mouse phenome data with > X sequenced strains, or a large gene expression dataset) was used.
Linked to this is a concern about power. Given the small number of strains (e.g. for interfrontal bone formation 9 strains) and the large number of variants (up to 74,873,854) how many variants would we expect to see segregating between the two groups by chance? A simple way of testing this would be to look at the number of variants called by your method in each combination of four and five strains, and how many of these meet the criteria that you used to prioritize genes for presence of the interfrontal bone. My worry is that you will get nearly as many candidate variants using a random split of strains as you got by splitting strains by phenotype.
These concerns are somewhat demonstrated in the authors’ Dystrophic cardiac calcification use case. In the previous QTL mapping, four loci were identified, whereas using the current ‘fine mapping’ method, 13 loci were identified. It is unclear to me how the authors prove that the additional 9 loci are not simply false positives.
Another issue is that only SNPs are used. Although these are easier to find in commonly used short-read sequencing data, it leaves all potentially larger variants (e.g. insertions and deletions), that could also alter the phenotypes, unaccounted for. In QTL mapping this is not such an issue, since the markers/SNPs will tag nearby large variants, but in this case, where variants (and therefore genes) are being prioritized based on their predicted effects then this is not the case.
Linked to this, variants are being prioritized based on their protein coding effects (e.g. ‘high’ impact). GWAS data in humans shows the importance of gene expression regulating variants. These variants would be missed by the current method. As a demonstration of this, one wonders how this method would perform in providing candidate variants for gene expression as a phenotype.
Finally, there does not appear to be any consideration of potential kinship between inbred strains. Given the relatedness among many inbred strains, it is possible that situations occur where strains share many variants simply due to this kinship. This becomes increasingly problematic as fewer strains are used for phenotyping. The inverse of this is that if closely related strains show different phenotypes, then there will be fewer variants between the two (and examples of this have been used for reduced complexity crosses). This effect of kinship should be mentioned.
Smaller, minor suggestions are below:
In the abstract, ‘In addition, lately, also’ does not read well, ‘In addition, recently’ would work better.
Line 35: ‘favourable animal model’ should be plural, ‘favourable animal models’
Lines 38-40. It would perhaps be better to refer to this strain as Black 6J. C57BL/6J, C57BL/6N, C57BL/6NJ and C57BL/6NCrl are all sometimes called Black 6, and indeed, you have C57BL/6N among your inbred strains.
Lines 47-50: This idea of in silico fine mapping has certainly been used since then, even if not using that exact phrase. Examples include Hall and Lammert, 2017 (10.1007/978-1-4939-6427-7_21) and Mulligan et al., 2019 (10.3389/fgene.2019.00188).
Line 109: End of parentheses missed
Lines 137: The sentence should start with ‘The’, i.e. ‘The function finemap takes’.
Line 149: “albinisim” should be “albinism”
Figure 4: Panel C is not mentioned in the legend.
Line 166: Should read ‘we applied our algorithm to the phenotype’
All text and materials provided via this peer-review history page are made available under a Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.