Information Theory for Biological Sequence Classification: A Novel Feature Extraction Technique Based on Tsallis Entropy
, Anderson P. Avila Santos 1,2, Breno L. S. de Almeida 1, Peter F. Stadler 3, Ulisses Nunes da Rocha 2,*, Danilo S. Sanches 4 and André C. P. L. F. de Carvalho 1Round 1
Reviewer 1 Report
This is a very excellent work elaborating on the utility of Tsallis and Shannon entropy for biological sequence classification. I only had a few comments:
Major comments:
1. This work introduced a novel and efficient method of biological sequence classification using information entropy. However, too less about biological meaning/interpretations/potential applications were addressed for this novel technique. I suggest adding a paragraph or at least a couple sentences for this method.
2. The Conclusion section is too lengthy -- it should be only a few sentences. I suggest moving some of the contents into the Discussion section.
Minor comments:
1. Line 64: "entropic index" not "Entropic index".
2. Line 139: "long non-coding" not "Long non-coding".
3. Line 148: Expand the abbreviation "SARS-CoV-2" as this was the first time in the text that this term was introduced.
4. Line 155: "assessed" not "assess".
5. Line 164 and 166: "used" not "use".
6. Line 167: "investigated" not "investigate".
I look forward to seeing a revised version of this work!
Author Response
We added our replies to the comments by reviewer 1 in the attached file.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
The authors use for comparison of k-mers histograms the Tsallis entropy instead of the more widely used Shannon Entropy. The experiments are extensive.
Unfortunately, I missed a bit the motivation why the Tsallis entropy and not any other entropy is used. The first part is very short and without new insides here. A more detailed comparison, motivation and also the consideration of other Entropy measurements would be very useful and improve the paper.
Moreover, a deeper literature research to the topic about alignment free comparison of sequences and a classification of the presented work is mandatory.
The experimental results are also ok, but not really show a significant improvement of the use of the Tsallis entropy. A comparison to other entropies and maybe also other information theory-based comparison methods of sequences would be give a better insight. The comparison with SVD/ UMAP I cannot understand, what has this to do with the presented work beside the use of kmers (here exists more sophisticated variants ==> literature research to state of the art and a comparison with this would be better, in my point of few)
Author Response
We replied to all comments by reviewer 2 in the attached file.
Author Response File: Author Response.pdf
