miGAP: miRNA–Gene Association Prediction Method Based on Deep Learning Model

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The prediction of MiRNA- gene association is quite important to identify many therapeutic targets and hence the potential methods to do this job should be identified and developed. In this manuscript the authors proposed a unique method called “deep learning approach” to predict associations between miRNAs and genes. This method was found to be outperformed the SOTA miRNA-gene association prediction methods. The outstanding performance of the proposed model attributed to Extensive Training Data, Optimal Data Embedding, Logical Negative Data Construction and Optimized Model Architecture.  

The two case studies conducted by them using their method seems reliable. The performed experiments to predict pairs with unknown associations proved to be reasonable. The experiments to predict miRNAs closely associated with BRCA2, a gene linked to various cancers was quite interesting. This deep learning approach clearly looks promising to have a wide range of practical applications in future. Over all, this work is worth to publish in Applied Sciences based on the current demand of interest.

Author Response

Dear Reviewer,

Thank you very much for your insightful review. We sincerely appreciate the time and effort you have taken to provide us with your valuable feedback. 

We look forward to making significant advancements in our work and contributing more to the field.

Warm regards,
Seungwon Yoon

Reviewer 2 Report

Comments and Suggestions for Authors

Authors describe a new approach to predict miRNA gene targets. As part of their study, they proposed a novel approach to create a negative (no interaction) sample set. The reported prediction performance is better than existing prediction tools.

 

The manuscript is easy to read, well-organized and complete. The methods are well described and illustrated. The results are reported on par with similar literature. The discussion probably needs some attention as its content can move to results.

 

My minor comments regarding the paper are:

 

-Paragrahs starting with "Inferring MiRNA...",  "LncMirNet [14] was developed..." and "The SG-LSTM-FRAME model"..  can be organized in alignment with Table 1. Starting these paragraphs with few introductory sentences describing tools' prediction target (e.g. gene, lncRNA..) will help the flow.

 

-biomaRt is mentioned as a gene data resource. However, Biomart is an interface to access/retrieve biological data from several resources. I believe authors must have used/accessed  Ensembl Gene or some other Gene resource through Biomart. So the text and citations need revision.

 

-In section 2.2 the introductory text on miRNA, gene, and ML can be removed or moved elsewhere (e.g. Introduction or Background) if the authors' intention is to provide more insight on the topic to general audience of the journal.

 

-The negative set generation as in other biological classification problems rightfully identified. Authors presented a negative sample selection methodology picking the most distant vector encoded miRNA-gene pairs to the positives using three different (euclidian, cosine, and mahalanobis) distances.  It is not clear why three distances are used collectively: An illustration such as Venn diagram showing the overlap/difference between negatives picked when individual distances are used could be supportive of their approach. 

 

-The choice of LSTM/Bi-LSTM rather than CNN or RNN is not justified well. Probably the last sentence in 2.4 is a start but it would be nice to extend it and move it to a prominent location in the flow (maybe the beginning of the same section)

 

-The subsections of discussion are actually can move to results. In Figure 6 one would expect to see PCA analysis on negatives/positives in one graph and somehow show separation (if any) from an angle if necessary. 

 

-In Table 10, miTars' accuracy can be reported as 0.955 and, for comparison purposes, miGAP's accuracy can be reported as well.

 

Author Response

Dear Reviewer,

We are immensely grateful for the thoughtful and detailed review you have provided. Your insights and suggestions have not only enhanced the quality of our work but have also shown us the care and attention you've dedicated to understanding and improving our research.

Reading through your comments, we felt a genuine sense of support and guidance which is truly heartening. It is clear to us that, thanks to your constructive feedback, our study has evolved and will be presented in a better light.

We sincerely appreciate the time and effort you invested in our work, and we hope that our revisions have met the high standards you hold for scientific inquiry.

Thank you once again for your invaluable contribution.

Warm regards,

Seungwon Yoon

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

The manuscript titled as "miGAP: MiRNA–Gene Association Prediction Method based on Deep Learning Model" aims to perform miRNA-target gene prediction, an area that has been studied extensively for the past years. 

The method/approach used in the study has been applied by others as it can be seen from table 1 and 10. The original outcome of the study is the newly generated negative data set. 

In the current form of the paper, there are various points that need to be improved:

- The text is not properly referenced.

- "Currently, there are approximately 2000 known species of miRNAs that play a key role in survival." They probably mean human miRNAs here, but it is not mentioned. 

- miRNA - target (gene) binding has different properties based on the organisms, i.e, humans and plants have a major difference. This issue is not considered in the study. 

- the quality of the graphs are poor. 

- the method and the data sets should be provided. 

Comments on the Quality of English Language

- The paper is difficult to follow. Language editing and a structured information flow are necessary. 

Author Response

Dear Reviewer,

We sincerely thank you for your insightful and thorough review. The points you raised were instrumental in guiding us towards significant improvements in our manuscript. Each comment you provided was a beacon that helped us sharpen our focus and enrich the clarity of our paper.

Please find attached a Word document in response to the comments provided. We have addressed each point in detail and have made corresponding revisions to the manuscript.

As we integrated your feedback, we observed our paper evolving, becoming more coherent and impactful. Your critical eye did not just critique; it illuminated paths for enhancement that we had not seen.

We are genuinely appreciative of the time and effort you invested in our work. It is reviewers like you who uphold the standard of excellence in scientific inquiry and propel progress through constructive criticism.

Warm regards,
Seungwon Yoon

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

The manuscript titled "miGAP: MiRNA–Gene Association Prediction Method based on Deep Learning Model" delves into the vital task of forecasting associations between microRNAs (miRNAs) and genes, which is pivotal for understanding their roles in molecular processes and advancing therapeutic development for diseases. This study aspires to surpass the performance of prevailing deep learning methods within this domain by confronting their limitations. To accomplish this, the authors curate a substantial embedded dataset encompassing 717,728 miRNA-gene pairs, meticulously tailored to suit their deep learning model. Additionally, they apply an embedding method originally devised for protein sequences to transform gene sequence data and construct a comprehensive negative dataset, contrasting with the random selection of negative data commonly employed in most studies. The ensuing deep learning model showcases superior performance compared to state-of-the-art methods, with an area under the receiver operating characteristic curve of 0.9834. The manuscript includes a case study demonstrating the model's predictive prowess and endeavors to identify miRNAs closely linked to genes associated with various cancers.

Reviewer Comments:

 

1. The manuscript should be examined for the following minor issues:
• Revise the Introduction section to ensure the accuracy of information related to miRNAs.
• Include clear figure legends for Figure 1, 3, 5, and 6. Reorganize Figure 2 for consistency in size and spacing. Confirm the format of Table 10. Please consider removing Supplementary Figures from the manuscript.
• Correct typos and inconsistencies in capitalization and italics when mentioning gene names throughout the manuscript.
2. In section 3.4 (Validation of the Model), it is recommended to:
• Include a more extensive list of target genes and apply statistical analyses to demonstrate the performance of the established model compared to existing models.
• Revise the section to eliminate irrelevant information not directly related to the results.
• Select target genes that exhibit mutations in the 3'UTR region, which may affect potential miRNA binding.
3. Discuss the factors that may influence the performance of the proposed model, and provide a detailed exploration of the advantages and limitations associated with this deep learning approach.
4. Revise the Conclusions section and consider moving the future plans (last paragraph) to the Discussion section for better thematic organization.
5. Clarify whether SOTA miRNA-gene association prediction methods are the sole platform currently in use, and elucidate the differences between the proposed deep learning model and existing methods.
6. Ensure that the Reference list adheres to a consistent format and contains all necessary information.

Comments on the Quality of English Language

Minor editing of English language required

Author Response

Dear Reviewer,

Thank you very much for your insightful review. We sincerely appreciate the time and effort you have taken to provide us with your valuable feedback. Inspired by your comments, we are committed to further developing this research through various innovative approaches.

Please find attached our "Response to Reviewer Comments" document, where we address each point you've raised in detail. We have taken great care to ensure that all of your feedback has been thoroughly considered and reflected upon in our revisions.

Thank you once again for your constructive comments, which have been invaluable to the enhancement of our work.

We look forward to making significant advancements in our work and contributing more to the field.

Warm regards,
Seungwon Yoon

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report

Comments and Suggestions for Authors

the authors covered majority of the issues mentioned in the previous review stage. 

Comments on the Quality of English Language

English is better than the previous version. 

Reviewer 4 Report

Comments and Suggestions for Authors

The manuscript has been improved with the authors' further efforts on major revisions.

I suggest that minor editing of the manuscript shall be conducted for the final proofreading.

Comments on the Quality of English Language

Minor editing of English language required