The value of a large Pisum SNP dataset

The authors have addressed all my suggestions, and I now recommend to accept the revised manuscript.

During the initial round of revisions for Aubert's research article entitled "SNP discovery by exome capture and resequencing in a pea genetic resource collection," I expressed a significant concern regarding the phylogenetic analyses, accompanied by several minor concerns. I am pleased to report that all of these concerns have been effectively addressed. The authors employed a coalescence-based phylogenetic inference in conjunction with the standard substitution model, leading to the observation of variations in the clades, which were attributed to adixture leading to incomplete lineage sorting (ILS).

Nevertheless, I would like to propose two final suggestions. 

1. Firstly, I would recommend the authors to incorporate clade support measures within their tree representations. While acknowledging that the authors are preparing a subsequent paper dedicated to comprehensive phylogenetic analyses, it is important to acknowledge that once a phylogenetic tree is published, it often becomes the foundation for subsequent analyses, even when certain clades may exhibit lower resolution. In light of the present study's indication of disparate outcomes between the two employed methods, likely due to ILS, providing measures of clade support derived from these methods becomes important. Such information would empower potential users of the analyses to assess the robustness of these discrepancies and make more informed decisions.

2. Additionally, I suggest highlighting the specification of the two methods used, namely the standard substitution model and the coalescence-based approach, in line 143.

Overall, I express my satisfaction with the revisions implemented in this manuscript.

Rui Borges (Institute of Population Genetics, Vetmeduni Vienna)

The paper by Aubert, "SNP discovery by exome capture and resequencing in a pea genetic resource collection," aims to evaluate the genetic diversity of a collection of 240 pea (Pisum sativum L.) accessions. This paper reports on the large number of SNPs (approximately 2.3 million) obtained through whole-exome sequencing. The methods described in the paper appear to be sufficiently detailed. Additionally, I agree with the authors' conclusions and recognize the potential value of this dataset for future genome-wide association studies and breeding programs in pea.

My main concern is with the phylogenetic analysis. The authors carried out their analysis using SNPs while ignoring constant sites (lines 92-94). This could potentially impact the estimated distances based on the GTR model. It is not clear to me if this will significantly affect the tree topology (although, I suspect it will not), but it could alter the branch lengths. I recommend that the authors take steps to correct for ascertainment bias in their analyses. Furthermore, the authors chose to present the tree as a cladogram, which does not allow the reader to appreciate the branch lengths. This information could be useful in understanding the differences found between the clustering and structure methods (lines 152-53). For example, the authors should consider whether long-branch attraction could be a factor.

My primary concern, however, is more fundamental. Given that the authors are working with genetic positions that are not fully sorted, can they accurately estimate a phylogenetic tree based on the GTR distance? I am uncertain how to interpret the estimated clades and distances in this context. Therefore, I believe it would be more appropriate to estimate a coalescent tree instead, as it provides a more appropriate description of the dataset the authors have on hand. This is of fundamental importance because I believe that some of the clades estimated in this work are likely to be used in further studies.

There are also a few minor aspects I would like to address:

• Lines 89-90: It would be helpful if the authors could explain why a 10% threshold was chosen for filtering SNPs.
• Lines 124-125: It would be beneficial to include a brief explanation of how the categories of low, moderate, and high were determined (this is only a suggestion).
• Line 127: Is it likely that the 0.53% SNPs with a disruptive nonsense effect (or total 0.2329*0.0053) could potentially be due to sequencing errors?
• Line 137: It is not clear to me how the clustering analyses separated the accessions according to crop evolution. Could the authors provide further clarification on this point?
• Lines 115-116: I found it peculiar that the cultivated winter pea fodder had only two singletons per accession. Is there a reason for this?
• Lines 152-153: It would be helpful if the authors could provide more specific reasoning for why they believe the differences in placement between the phylogenetic analyses and the structure/clustering analyses are due to kinship.

Rui Borges (Institute of Population Genetics, Vetmeduni Vienna)

In the manuscript “SNP discovery by exome capture and resequencing in a pea genetic resource collection”, the authors performed exome sequencing to genotype 240 accessions of pea and identified a dataset of SNPs to be used for genetic diversity analysis. For this, the authors selected a large number of samples, including cultivars, landraces, and wild types with diverse geographical origins that follow the standard approach of population genetic analysis. This study is interesting and has valuable information, but some details still need to be improved. The following are some questions and suggestions for modification:

• I wish the authors could state why exome-derived SNPs were chosen instead of GBS SNPs, which are widely used to assess genetic diversity in several plant species, including those with complex genomes.
• In the background section, the authors should add the genome characteristics data: number of chromosomes, ploidy level.
• If possible, I suggest the authors indicate the sample locations in the map figure, so it will be easier to see the geographical distribution.

  ---

• In the method section, the author should provide information about the number of probes used in this study.
• The authors should provide more details about the criteria used to select the subset of SNPs. Do you filter SNPs based on MAF? What about the threshold for linkage disequilibrium?
• I suggest the authors add an analysis of maker polymorphism (PIC), genetic diversity parameters, as well as genetic differentiation (FST) among sub-populations.
• The authors used a maximum likelihood phylogenetic tree, according to lines 94–95, but neighbor-joining phylogenetic trees are mentioned in line 134.What is the method used for the analysis of the phylogenetic tree?
• For the structure analysis, the authors should also provide information regarding the settings and parameters for software used. What threshold value of the membership coefficient was used to assign an accession to a specific group or assign an accession as admixture?

  ---

• In the results section, the authors should report the number of raw reads obtained from sequencing, the %map read with the reference genome, and the number of raw initial SNPs obtained.
• The authors analyzed k values from 1 to 10. They should show a plot or the statistic that indicates which is the best value of K.
• Line 137: In my version, it does not have table 1.
• Figure1: If possible, I suggest the authors separate the DAPC plot into Figure 2 and color the branches of the phylogenetic tree according to crop evolution and cultivation types, as well as include bootstrapping supporting values in the tree.

  ---

• The manuscript should have a discussion section and should be interpreted with the results as well as discussed in relation to the present literature.