Odate for the remark of the reviewer, we moved the description
Odate for the remark of the reviewer, we moved the description of the weighting strategies to the supplementary text (Additional file 1: Text S1). Line 261: “For weighted voting, ranking is done after collecting the weighted votes (see the detailed description in Additional file 1: Text S1)”. We moved the following section from the methods section to Additional file 1: Text S1: “The weighted votes are collected as follows: Suppose a combination of methods is denoted by , || is the number of methods in the method combination , and wi is the weight of sequence-based method Mi contributed to the coevolution scores. All methods contribute equally when every wi equals to 1. The normal??ized coevolution scores C ?; Dj is defined as: n;m ????1 X C ?; Dj ?wi ?C ?M PubMed ID:https://www.ncbi.nlm.nih.gov/pubmed/28607003 i ; Dj n;m n;m jj Mi ??C ?; Dj is thereafter ranked and exported as outn;m puts. Notably, can either contain a single method or a combination of methods, which can be selected based on the performance evaluation”. Comment 4: The strength of the methods used is the inclusion and comparison of a PubMed ID:https://www.ncbi.nlm.nih.gov/pubmed/29069523 number of different algorithms. Most of the weaknesses are presented properly and are generally applicable to similar coevolutionary studies. The manuscript is supported by rich supplementary data, which is important for the understanding and reproducibility of the results. I especially welcome that the authors provide the MATLAB source code for their ensemble method. The writing and organization of the manuscript is clear, the figures are appropriate and informative, although I think that Figures 3 and 4 could go as supplementary material, especially as the data presented in Figure 3A are closely related to those shown in Table 1. I would suggest that the authors consider the following points for preparing the final version of the manuscript: Author’s response: We thank the reviewer for this suggestion. We get SCR7 rearranged Figure 3A to Additional file 1: Figure S1. As for Figure 4, we consider it as a key representation of our clustering analysis which shows that similar sequence-based methods also provide similar predictions. While these findings could be perceived as seemingly trivial, we do believe that our visual representation is conceptually important. Comment 5: I would have desired a short outlook with an alternative data set (e.g. a single protein from a wider range of HIV1 subtypes, as mentioned by the authors asLi et al. Biology Direct (2015) 10:Page 17 ofa potential source of conflict with results of an independent study) and the discussion of an example where coevolving residues are clearly not in spatial contact. Author’s response: Indeed, our discussion mentioned that: “Our study observed different predictions within matrix and capsid, possibly because we focused on HIV-1 subtype B, while the coevolution analysis in [5] used a mixed subtype B and C dataset. Further investigation needs to distinguish coevolving residues in HIV-1 subtypes B and C”. We would like to give more details to explain this difference between different HIV-1 subtypes, in addition to our recent publications exploring such differences [57,86]. Currently, there are 8 HIV-1 subtypes and more than 60 classified circulating recombinant forms (CRFs) recorded in the Los Alamos National Laboratory (LANL) database (http://www.hiv.lanl.gov/content/index). The amino acid sequence diversities of HIV-1 Gag proteins between different subtypes and CRFs are between 15 and 20 [86]. More importantly, 103 (20.6 ) of 500 Gag amin.
