Machine Learning-Based Models for Detection of Biomarkers of Autoimmune Diseases by Fragmentation and Analysis of miRNA Sequences
Round 1
Reviewer 1 Report
The work entitled "Machine Learning-Based Models for Biomarkers Detection of Autoimmune Diseases by Fragmentation and Analysis of miRNA Sequences" aims to present two complete models for biomarker detection for two autoimmune diseases, multiple sclerosis and rheumatoid arthritis, using miRNA analysis. The two models presented are based on work previously published by the authors and include complete pipelines of text mining methods integrated with traditional machine learning and LSTM deep learning methods. The paper is well written and has the makings of a publication.
Author Response
Dear Reviewer,
Thank you so much for reviewing our work, we really appreciate your time and efforts. Please see the attachment.
Sincerely,
Nehal Ali
Author Response File: Author Response.docx
Reviewer 2 Report
Nice study but in some areas, it is poorly written.
Grammar needs to be checked and proofread before submission.
Figure numbers are confusing. After figure 8, figure 7 comes back and the second sequence begins.
Figure legends need to be checked for grammar and proofreading.
Please try to show and explain statistics in figure legends. The legends are very concise and insufficient description of the figure.
Overall it might be difficult for the reader to follow the manuscript.
Author Response
Dear Reviewer,
Thank you so much for reviewing our work, we really appreciate your time and efforts. Please see the attachment.
* Please note that the line numbering is impacted by applying the "track changes" feature
Sincerely,
Nehal Ali
Author Response File: Author Response.docx
Reviewer 3 Report
In the paper "Machine Learning-Based Models for Biomarkers Detection of
Autoimmune Diseases by Fragmentation and Analysis of miRNA Sequences" the authors presented new machine learning-based models for better detecting miRNAs as biomarkers. The paper is intensely focused on the mathematical background of the modeling process. However, considering that this paper is dealing with models for better biomarker discovery, the manuscript would greatly benefit (especially in the discussion) from simplifying certain technical aspects and clarifying in a manner that the audience not familiar with the machine learning can understand the process. I suggest that during the discussion the authors give more concrete conclusions about every result obtained, and give more comments on how this can be practically exploited. For example, it is hard to understand what library prep approach (NEBNEXT and NEXTFLIX) is better (based on the obtained results from proposed models). In addition, I am not sure why the authors have focused on MS and RA? They have stated in the limitations that larger data sets used are limited by the number of samples. Why they have not tried their models on some other diseases with larger sample sizes? Finally, I was wondering how the authors can be sure that these models can improve biomarker detection. They have compared the new models to the old ones (also designed by the authors), but maybe it would be also good to compare them to some other models that were designed on the larger datasets.
Minor comments:
Hence, Different sequencing preparation protocols involve studying their influence 62 on the sequenced data and how these various protocols impact the diseases’ biomarkers 63 detection accuracy and the results of genetic studies[10][11]. - correct Different to different
Figure 14. the obtained accuracy scores of Ex06, Ex07, and Ex08. - correct the to The
Author Response
Dear Reviewer,
Thank you so much for reviewing our work, we really appreciate your time and efforts. Please see the attachment.
*Please note that the line numbering is impacted by enabling the "track changes" feature in Word.
Sincerely,
Nehal Ali
Author Response File: Author Response.docx