Stochastic Safety Radius on UPGMA

Round 1

Reviewer 1 Report

algorithms-2064609

Yoshida et al., submitted an article manuscript entitled “Stochastic Safety Radius on UPGMA “for publication consideration in Algorithms.

The report provided unweighted Pair Group Method with Arithmetic Mean (UPGMA) which is considered as a popular distance-based methods to reconstruct a phylogenetic tree from a distance matrix computed from an alignment of sequences. Where The method of Stochastic safety radius was introduced by Steel and Gascuel. 

The scientific soundness of this manuscript is acceptable. It meets the aims and scope of algorithms journal. 

I think this report is sounds. The manuscript is well written. However, the content and substance of article is brief, it may fall into as communication report. 

It would be beneficial to reader if authors could address some minor points listed below: 

 (1) It is essential to supplement the comparison between theorem 10 and theorem 5 and explain the advantages of theorem 10. It would be better to give a formal proof of theorem 10;

(2) The experimental data is not well explained in the experimental part. It would be appreciated to provide more the experimental data examples here which could further support the claims of the application effect on the algorithm.

In my opinion, this manuscript is well written and in very sound presentation if reclassify as communication for type.  

Author Response

Thank you so much for very positive comments.  You insight is appreciated and has helped to improve our submission.

Comment #1: (1) It is essential to supplement the comparison between theorem 10 and theorem 5 and explain the advantages of theorem 10. It would be better to give a formal proof of theorem 10;

Response: Thank you for this insight.  Adding computational experiments for n = 3 allows us to compare Theorem 10 and Theorem 5 in Section 4.  Please see the new Figure 5 for with these results.  
For the proof, the interest in additional support is appreciated. Please see the formal proof for Theorem 10 to include equation (18)  at the end of section 3.

Comment #2: (2) The experimental data is not well explained in the experimental part. It would be appreciated to provide more the experimental data examples here which could further support the claims of the application effect on the algorithm.

Response: The identification of the value of more experimental data examples is a very good point. We address this through inclusion of additional experimental results and thorough discussion in Introduction and Computational experiments.  We added experimental results for explaining your comment (1). 

Author Response File: Author Response.pdf

Reviewer 2 Report

Review of the article "Stochastic Safety Radius on UPGMA".

The UPGMA algorithm assumes that the tree is additive and ultrametric, i.e. all taxa are at the same distance from the root - this assumption is usually incorrect.

The same, most often incorrect, assumption is used to root the UPGMA tree. For this reason, UPGMA is rarely used by phylogeneticists today and is discouraged for the construction of phylogenetic trees.

For this reason, the Authors should thoroughly explain why this unpopular and ineffective method became the subject of the article.

Author Response

Referee #2

Thank you so much for your comment.  Please see our address below.

Comment #3: The UPGMA algorithm assumes that the tree is additive and ultrametric, i.e. all taxa are at the same distance from the root - this assumption is usually incorrect.
The same, most often incorrect, assumption is used to root the UPGMA tree. For this reason, UPGMA is rarely used by phylogeneticists today and is discouraged for the construction of phylogenetic trees.
For this reason, the Authors should thoroughly explain why this unpopular and ineffective method became the subject of the article.

Response:  The first author, familiar with methodological  challenges of UPGMA and ultra metrics, appreciates this point .  This author has published over 80 papers over the last 15 years in the domain of phylogeny and phylogenomics and is excited the reviewer is keenly aware of tradeoffs in methodological approach.  Our approach is also motivated by this point. The address, and this paper, is encouraged by phylogenomics, a new field which applies tools in phylogenomics to genome datasets.  In phylogenomics, the multi species coalescent model is often used for the species tree gene tree analyses.  Under this model, we assume that all gene trees are ultra metrics (or equidistant trees).  Since we focus on phylogenomics, we focus on equidistant trees.  It is clear this approach and foundation was not clear; we have added some recent works on gene trees and species tree analyses in Introduction section.

Author Response File: Author Response.pdf

Reviewer 3 Report

In this study, the authors proposed a stochastic safety radius computation of UPGMA for a phylogenetic tree with n leaves. The idea seems interesting, however, some major points should be addressed as follows:

1. The review of related work is not sufficiently thorough and not sufficiently specific.

2. The authors only mentioned the results while there is no in-depth discussion.

3. The authors are suggested to update "References" by including the latest publications in the related work.

4. Abstract should be improved by adding more information.

5. Quality of figures should be improved.

Author Response

Referee #3

Thank you so much for your comments.  Please see below for our response to each comment.

Comment #4: 1. The review of related work is not sufficiently thorough and not sufficiently specific.

Response:  We appreciate the opportunity to further motivate our work with additional references. We added more to identify the extensive work establishing space of equidistant trees, and the work investigating this limiting assumption through application of polyhedral geometry.

Comment #5: 2. The authors only mentioned the results while there is no in-depth discussion.

Response:  The identification of a need for greater discussion is a very good point. We address this through the addition of additional computational results and thorough discussion in Introduction and Computational experiments. 

Comment #6: 3. The authors are suggested to update "References" by including the latest publications in the related work.

Response:  We appreciate the opportunity to further motivate our work with additional references. We added more to identify the extensive work establishing space of equidistant trees, and the work investigating this limiting assumption through application of polyhedral geometry.

Comment #7: 4. Abstract should be improved by adding more information.

Response:  The abstract was light on defining the specific problem we addressed and implications of these results to the greater field of phylogenomics. This is a very good point and we appreciate the opportunity to address these important perspectives.

Comment #8: 5. Quality of figures should be improved.

Response:  The pictures were increased in size, reducing surrounding whitespace and improving interpretation.  This is a very good point and we appreciate the opportunity to show these results in greater detail.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

This is research relevant to genetics. Authors are familiar with machine learning ideas. 

Reviewer 2 Report

Thank you for the correction. Before the correction it was really hard to understand the meaning of the article.

All my comments and concerns have been properly taken into account in the article. Now the article is much better and in my opinion it can be published in Algorithms.

Reviewer 3 Report

My previous comments have been addressed.