Review Reports - Exploring the Role of Extracellular Vesicles in Pancreatic and Hepatobiliary Cancers: Advances Through Artificial Intelligence

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This review article (Manuscript ID: ijms-4106811) introduces first the biology and oncological role of extracellular vesicles (EVs), next the sources, cargos, and roles of EVs in pancreatic and hepatobiliary cancers, then the AI-assisted EV research and the challenges. At present, both EVs and AI are hot research topics and the AI-assisted EV research is an interdisciplinary, rapidly growing direction of EV research. I have the following comments.

Major comments:

Figures should be added to summarize many parts of the manuscript for the sake of readers to understand the main contents of the article rapidly.
The descriptions in the main text overlap the main contents of Tables (e.g., lines 223-345 for Table 1, lines 376-441 for Table 2, lines 484-607 for Table 3, lines 642-800 for Table 4). (a) It is unnecessary to repeat the content. The main text can describe some contents which are not covered by the table, or describe the most important parts of the table in details, or summarize their main similarities and differences. (b) The authors should organize the descriptions in the main text in a form of paragraphs instead of in an item-listing form similar to a table. (c) The references should be cited in the corresponding locations of the Table. (d) By the way, connecting EV and its cargo in a word (e.g., EV-CD147, EV-miR-92a-3p, EV-miR-21, etc.) seems not a regular expression in publications.
It seems unnecessary to introduce EV isolation and detection methods which are not related with the topic of this review article. If the authors want to keep this subsection, it should be shortened/concise and moved to the end of section 2, e.g., as subsection 2.4. Moreover, references should be added at the corresponding locations in Table 5 (take the contamination with lipoproteins for example, please refer to “Lan, M., et al. Solving the contamination conundrum derived from coisolation of extracellular vesicles and lipoproteins: Approaches for isolation and characterization. Small Methods. 2025. 9(11): e01606” and “Simonsen, B., What are we looking at? Extracellular vesicles, lipoproteins, or both? Circulation Research. 2017. 121: 920-922”).
The AI parts (i.e., sections 5 & 6) of the article introduce too many backgrounds of AI, and the introduction about AI-assisted/mediated EV research involves a broad range of cancers instead of focusing on pancreatic and hepatobiliary cancers, which does not fit with the manuscript title (“Towards innovative approaches in pancreatic and hepatobiliary cancer”).

Minor comments:

Sections 1 & 2: It would be better to use the nomenclature of EV (or EV subtypes) recommended recently in MISEV 2023 by International Society for Extracellular Vesicles (ISEV). Please refer to “Zhang, Y., et al. Minimal information for studies of extracellular vesicles (MISEV): Ten-year evolution (2014-2023). Pharmaceutics. 2024. 16(11): 1394” and “Welsh, J.A., et al. Minimal information for studies of extracellular vesicles (MISEV2023): From basic to advanced approaches. Journal of Extracellular Vesicles. 2024. 13: e12404”.
The abbreviations below the Tables should be listed in an alphabetic order of the first letter of the abbreviations.
Lines 1111-1148: Why are those words/phrases highlighted in bold?
Table 6 is missing, or Table 7 should be Table 6.

Author Response

Response to Reviewer
Reviewer 1
This review article (Manuscript ID: ijms-4106811) introduces first the biology and
oncological role of extracellular vesicles (EVs), next the sources, cargos, and roles of EVs
in pancreatic and hepatobiliary cancers, then the AI-assisted EV research and the
challenges. At present, both EVs and AI are hot research topics and the AI-assisted EV
research is an interdisciplinary, rapidly growing direction of EV research. I have the
following comments.,
Reply: We sincerely appreciate the time and effort you and the reviewers have dedicated
to evaluating our manuscript. We have carefully considered the concerns raised and
have revised our manuscript accordingly. Below, we provide a point-by-point response to
each issue, outlining the modifications made or clarifications provided. Additionally, we
have substantially rewritten Sections 5 and 6 to address the reviewer’s concerns,
refocusing the content on EV-based AI applications in pancreatic and hepatobiliary
cancers, strengthening the conceptual motivation for AI, and adding critical discussion
of limitations and translational challenges.
Major comments:
Comment 1. Figures should be added to summarize many parts of the manuscript for the
sake of the readers to understand the main contents of the article rapidly.
Reply: The manuscript has been revised accordingly. Please see updated Sections 5–8,
particularly Section 7 and Table 6. We have added several figures summarizing the EV
isolation and detection methods along with their limitations. The text has been
streamlined, with detailed information now provided only in the figure legends and
accompanying table.
Comment 2: The descriptions in the main text overlap the main contents of Tables(e.g.,
lines 223-345 for Table 1, lines 376-441 for Table 2, lines 484-607 for Table 3, lines 642
800 for Table 4). (a) It is unnecessary to repeat the content. The main text can describe
some contents which are not covered by the table, or describe the most important parts
of the table in details, or summarize their main similarities and differences. (b) The
authors should organize the descriptions in the main text in a form of paragraphs instead
of in an item-listing form similar to a table. (c) The references should be cited in the
corresponding locations of the Table. (d) By the way, connecting EV and its cargo in a
word (e.g., EV-CD147, EV-miR-92a-3p, EV-miR-21, etc.) seems not a regular expression
in publications.
Reply: We modified Sections 1 and 2 accordingly. The text was reduced, with key EV
subpopulations described in the main paragraphs, while additional EVs are summarized
in Table 1. Throughout the revised manuscript, EVs and their cargoes are consistently
referred to as EVs carrying specific molecules (e.g., EVs carrying miR-21).
Comment 3. It seems unnecessary to introduce EV isolation and detection methods
which are not related with the topic of this review article. If the authors want to keep this
subsection, it should be shortened/concise and moved to the end of section 2, e.g., as
subsection 2.4. Moreover, references should be added at the corresponding locations in
Table 5 (take the contamination with lipoproteins for example, please refer to “Lan, M.,
et al. Solving the contamination conundrum derived from coisolation of extracellular
vesicles and lipoproteins: Approaches for isolation and characterization. Small Methods.
2025. 9(11): e01606” and “Simonsen, B., What are we looking at? Extracellular vesicles,
lipoproteins, or both? Circulation Research. 2017. 121: 920-922”).
Reply: Response: Thank you for your comment. We have removed this section and
incorporated the key information regarding the limitations of each isolation and detection
method into the figure legend and Table 5. In addition, we have strengthened Section 7
(revised numbering), Challenges for AI in EV-Based Oncology. We also added both of the
Comment 4: The AI parts (i.e., sections 5 & 6) of the article introduce too many
backgrounds of AI, and the introduction about AI-assisted/mediated EV research
involves a broad range of cancers instead of focusing on pancreatic and hepatobiliary
cancers, which does not fit with the manuscript title (“Towards innovative approaches in
pancreatic and hepatobiliary cancer”).
Reply: The manuscript has been revised accordingly. Please see updated Sections 5–8,
particularly Section 7 and Table 6. We have substantially revised Sections 5 and 6 to
focus explicitly on pancreatic and hepatobiliary cancers and EV-centered AI
applications. Generic AI background has been condensed and refocused on EV-specific
modeling challenges (high dimensionality, non-linearity, and multimodal integration).
Broad cancer examples were either removed or reframed to illustrate methodological
principles directly relevant to pancreatic and hepatobiliary malignancies. We also
introduced explicit critical appraisal of study limitations and translational gaps.
Minor comments:
• Sections 1 & 2: It would be better to use the nomenclature of EV (or EV
subtypes) recommended recently in MISEV 2023 by International Society for
Extracellular Vesicles (ISEV). Please refer to “Zhang, Y., et al. Minimal information
for studies of extracellular vesicles (MISEV): Ten-year evolution (2014-2023).
Pharmaceutics. 2024. 16(11): 1394” and “Welsh, J.A., et al. Minimal information
for studies of extracellular vesicles (MISEV2023): From basic to advanced
approaches. Journal of Extracellular Vesicles. 2024. 13: e12404”.
Reply: Thank you we made the modifications that you suggested, and we added the
recommended citation
• The abbreviations below the Tables should be listed in an alphabetic order of
the first letter of the abbreviations.
Reply: Thank you, we made the modification that you suggested.
• Lines 1111-1148: Why are those words/phrases highlighted in bold?
Reply: Thank you. We removed the highlighting, which was added by omission.
• Table 6 is missing, or Table 7 should be Table 6.
Reply: We modified the tables and the numerical order of them.
.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

This comprehensive review manuscript provides a timely and detailed synthesis of the role of extracellular vesicles (EVs) in pancreatic and hepatobiliary cancers, with a particular focus on how artificial intelligence (AI) is accelerating biomarker discovery and therapeutic innovation. The work is well-structured, evidence-rich, and addresses a rapidly evolving field at the intersection of oncology, nanotechnology, and computational biology. The authors successfully highlight the translational potential of EV-based liquid biopsy and AI-driven analytics. However, several areas require clarification, reorganization, and methodological strengthening before publication.

Comments

While the content is exhaustive, the subsections (HCC, GBC, CCA, PDAC) are overly dense with lists of EV cargoes and effects, which may overwhelm the reader. Summarize key EV biomarkers in concise tables earlier in each subsection, then elaborate on mechanistic insights and clinical relevance in the text. The current tables (e.g., Table 1 for HCC) are helpful but appear after lengthy enumerations.
The transition from EV biology (Sections 2–4) to AI applications (Section 5) feels abrupt. The link between EV heterogeneity and AI’s ability to model such complexity should be emphasized earlier. Add a brief introductory paragraph at the start of Section 5 explicitly stating why AI is uniquely suited to address challenges in EV research (e.g., high-dimensional omics data, non-linear biomarker interactions).
The review includes many AI-EV studies but does not consistently critique their limitations (e.g., small sample sizes, lack of external validation, spectrum bias). Incorporate a dedicated paragraph or table summarizing common pitfalls in AI-EV research (e.g., overfitting, dataset bias, reproducibility issues) and how future studies might address them.
While EV isolation methods and challenges are discussed, the link between technical variability and AI model performance is underdeveloped. Expand Section 6.1 to include specific examples of how batch effects in EV isolation impact AI model training and generalization.

Some tables (e.g., Table 5 on EV isolation methods) are informative but could be better formatted for readability. Use consistent formatting across tables, include a column for “clinical applicability” where relevant, and consider moving detailed cargo lists to supplementary materials.
Abbreviations are well-defined, but some terms (e.g., large oncosomes vs. oncosomes) are used interchangeably in early sections. Clarify distinctions upfront..
The “Future Directions” section is brief. Given the rapid evolution of AI and EV therapeutics, a more forward-looking perspective—e.g., on federated learning, real-time monitoring, or regulatory pathways—would strengthen the conclusion. Expand with 2–3 paragraphs on emerging trends (e.g., AI-designed EV mimetics, integration with digital pathology).

Author Response

Response to Reviewer
Reviewer 2
This comprehensive review manuscript provides a timely and detailed synthesis of the role
of extracellular vesicles (EVs) in pancreatic and hepatobiliary cancers, with a particular
focus on how artificial intelligence (AI) is accelerating biomarker discovery and therapeutic
innovation. The work is well-structured, evidence-rich, and addresses a rapidly evolving
field at the intersection of oncology, nanotechnology, and computational biology. The
authors successfully highlight the translational potential of EV-based liquid biopsy and AI
driven analytics. However, several areas require clarification, reorganization, and
methodological strengthening before publication.
Reply: We sincerely appreciate the time and effort you and the reviewers have dedicated to
evaluating our manuscript. We have carefully considered the concerns raised and have
revised our manuscript accordingly. Below, we provide a point-by-point response to each
issue, outlining the modifications made or clarifications provided. Additionally, we have
substantially rewritten Sections 5 and 6 to address the reviewer’s concerns, refocusing the
content on EV-based AI applications in pancreatic and hepatobiliary cancers, strengthening
the conceptual motivation for AI, and adding critical discussion of limitations and
translational challenges.
Comment 1: While the content is exhaustive, the subsections (HCC, GBC, CCA, PDAC) are
overly dense with lists of EV cargoes and effects, which may overwhelm the reader.
Summarize key EV biomarkers in concise tables earlier in each subsection, then elaborate
on mechanistic insights and clinical relevance in the text. The current tables (e.g., Table 1
for HCC) are helpful but appear after lengthy enumerations.
Reply: Thank you for your comments. We extensively modified the skeleton of this
manuscript in order to be more easily read. We modified the positions of the tables.
Comment 2: The transition from EV biology (Sections 2–4) to AI applications (Section 5)
feels abrupt. The link between EV heterogeneity and AI’s ability to model such complexity
should be emphasized earlier. Add a brief introductory paragraph at the start of Section 5
explicitly stating why AI is uniquely suited to address challenges in EV research (e.g., high
dimensional omics data, non-linear biomarker interactions).
Reply: We extensively modified the skeleton of this manuscript. We added a new
introductory paragraph at the beginning of the section explaining how EV heterogeneity,
high-dimensional omics data, and non-linear biomarker interactions necessitate AI-based
modeling approaches.
Comment 3: The review includes many AI-EV studies but does not consistently critique
their limitations (e.g., small sample sizes, lack of external validation, spectrum bias).
Incorporate a dedicated paragraph or table summarizing common pitfalls in AI-EV research
(e.g., overfitting, dataset bias, reproducibility issues) and how future studies might address
them.
Reply: The manuscript has been revised accordingly. Please see updated Sections 5–8,
particularly Section 7 and Table 6. We added explicit methodological critique throughout
Section 7 (e.g., small cohorts, lack of external validation, spectrum bias) and consolidated
these issues, with emphasis on overfitting, reproducibility, and dataset bias. We also added
a table as you suggested.
Comment 5: While EV isolation methods and challenges are discussed, the link between
technical variability and AI model performance is underdeveloped. Expand Section 6.1 to
include specific examples of how batch effects in EV isolation impact AI model training and
generalization.
Reply: Section 7 ( previously Section 6) was expanded to explicitly discuss how batch
effects, isolation protocol variability, and platform-dependent biases impair AI training and
generalization, and how harmonization and MISEV-aligned reporting mitigate this.
Comment 6: Some tables (e.g., Table 5 on EV isolation methods) are informative but could
be better formatted for readability. Use consistent formatting across tables, include a
column for “clinical applicability” where relevant, and consider moving detailed cargo lists
to supplementary materials.
Reply: We have modified the tables as suggested and added several figures to improve
readability. The tables detailing extracellular vesicle cargoes have been revised, and the
paragraph formats have been condensed for clarity and ease of reading. The title was also
updated to emphasize the critical role of extracellular vesicles in these specific malignancies
and to highlight the importance of implementing artificial intelligence in this field. For this
reason, we have chosen to keep the extracellular vesicle cargoes and their effects in the
main text rather than moving them to the supplementary material.
Comment 7: Abbreviations are well-defined, but some terms (e.g., large oncosomes vs.
oncosomes) are used interchangeably in early sections. Clarify distinctions upfront.
Reply: We made the modifications that you suggested.
Comment 8: The “Future Directions” section is brief. Given the rapid evolution of AI and EV
therapeutics, a more forward-looking perspective—e.g., on federated learning, real-time
monitoring, or regulatory pathways—would strengthen the conclusion. Expand with 2–3
paragraphs on emerging trends (e.g., AI-designed EV mimetics, integration with digital
pathology).
Reply: We made the modifications that you suggested. Section 6. (on revised manuscript)
AI-assisted EV Therapeutic Engineering & Drug Discovery Please see updated Sections 5–8,
particularly Section 7 and Table 6. The part you suggested was expanded to include
multimodal foundation models, federated learning, generative EV cargo design, and real
time EV biomarkers, providing a forward-looking translational roadmap. Additionally, we
expanded section 8 ( on revised manuscript) Emerging AI-assisted EV research
methodologies and future perspectives

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The revised manuscript has addressed most of my previous concerns. I respect the authors' efforts. Honestly speaking, however, I do not like this article because it is too lengthy (not concise) and wants to include all things. I still have the following minor comments: (a) The font of the words in some figures including the graphical abstract is too small to be recognized easily; (b) there are two Figure 2, and the second Figure 2 should be Figure 5; (c) it seems that the images in Figure 4 and in the second Figure 2 are displayed in duplicate.

Author Response

We sincerely thank the reviewer for their careful re-evaluation of the revised manuscript and for acknowledging our efforts in addressing the previous concerns. We appreciate the reviewer’s candid feedback regarding the overall length and scope of the article. While we aimed to provide a comprehensive treatment of the topic, we have made concerted efforts to minimize and streamline the majority of the sections.

The manuscript is intentionally structured to accommodate readerships that may not be fully familiar with either extracellular vesicle (EV) biogenesis or artificial intelligence (AI), including its applications and limitations. In this context, the consolidation of key EV-related information into tables is intended to enhance clarity and accessibility, particularly for readers exploring these fields in relation to biomarker discovery.

Regarding the specific minor comments:

(a) We agree that the font size in some figures, including the graphical abstract, is currently too small. All affected figures will be revised to increase font size and improve readability.

(b) We thank the reviewer for raising the concern regarding potential duplication in figure numbering. Upon careful re-examination of the manuscript, we confirm that the figures are correctly numbered and sequentially ordered, as outlined below:

Figure 1: EV biogenesis and its role in intercellular communication within the tumor microenvironment (TME).
Figure 2: Overview of AI computational methodologies and their implementation in EV-related research.
Figure 3: AI applications in EV research in pancreatic and hepatobiliary malignancies.
Figure 4: AI-assisted EV therapeutic engineering and drug discovery.
Figure 5: EV isolation and detection methods and limitations of EV utilization.

We have also reviewed a clean version of the manuscript (with all track changes removed) and did not identify any duplicated figures or numbering inconsistencies. It is possible that the perceived duplication arose from the track-changes view in a previous version; nevertheless, we have rechecked the manuscript carefully to ensure clarity and consistency.

We remain grateful for the reviewer’s careful attention to detail and would be happy to address any remaining concerns.

The second instance of Figure 2 is indeed an error and will be corrected to Figure 5 in the revised version.

(c) We acknowledge the reviewer’s observation concerning the apparent duplication of images in Figure 4 and the second Figure 2. We will carefully review these figures and ensure that each image is correctly presented and clearly differentiated, correcting any inadvertent duplication.

We are grateful for these constructive comments, which will help us further improve the quality and clarity of the manuscript.