Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

A significant amount of work is evident in the collection, processing, and integration of data. The database includes 699 samples and over 7.47 million predicted CCC interactions. The structure of the service is clear, the navigation is intuitive, and the ability to analyze your own data is a valuable feature.

The appearance of the website: Despite the good organization, the interface requires improvement. The font and the large number of intense, heterogeneous colors distract and do not correspond to the character of a scientific database. I suggest using a more restrained and consistent visual design and a color palette that takes accessibility and color blindness into account.

Performance: During testing, there were very long delays, especially when loading interactive maps/visualizations. This is important because the authors present the ability to interactively browse millions of interactions as a key CCCView functionality. In the Methods section, they mention Django + SQLite. The authors should provide information on the infrastructure, bandwidth, typical response times, and performance when multiple users access the service simultaneously. It would also be worth clarifying whether SQLite is actually used in the production service.

User data upload: The function is useful, but the authors should provide information about its limitations (e.g., maximum data size), typical analysis time, computational resource limitations, and task queuing methods. It is also important to explain the rules for storing and deleting uploaded data and how data confidentiality is ensured.

Heterogeneity of the 699 samples: A unified pipeline facilitates comparisons, but the data originate from different technologies and have different spatial resolutions and sequencing depths (which the authors also acknowledge as a limitation). At the same time, fixed preprocessing parameters are applied across the datasets. A better justification of this strategy and a discussion of its potential impact on comparisons between samples would be useful in the manuscript.

CCC parameters: The fixed dis_thr = 500 μm used for COMMOT is particularly worth explaining. For data originating from different ST platforms, the authors should justify why a single value is appropriate regardless of technology, spatial resolution, and tissue type.

Cell-type annotation: Since Cell BLAST/Tangram are used to generate unified annotations across the entire database, it would be useful to demonstrate their reliability using a representative subset of the datasets, for example by comparing the resulting annotations with the original annotations provided by the authors of the source datasets. The glioblastoma case study illustrates why this is important—for example, annotations such as Purkinje cell appear in the results, while the authors themselves note that these represent transcriptomic assignments rather than definitive anatomical identification.

MuCIA: This is one of the most original components of the study and therefore deserves more convincing validation. The authors present the Tweedie model, Wald test, parameter estimation, and analysis of interaction distributions. If possible, it would be useful to provide a benchmark or simulation demonstrating the behavior of the method when the true group effect is known. The manuscript should also provide a clear explanation of multiple-testing correction, the handling of zero values, and precisely which outputs from the individual CCC methods are used as input for MuCIA.

Reproducibility/provenance: For a resource of this scale, a comprehensive metadata table would be useful, allowing each sample to be traced back to the original study/accession and providing information such as the platform, species, tissue, condition, etc. The authors should also clarify database versioning, the planned update strategy, and the availability of the analysis pipeline code.

The main value of this study lies in the scale and integration of the resource and in the potential usefulness of CCCView. Before publication, however, several aspects would benefit from further improvement or clarification, particularly the performance and reproducibility of the platform, the validation of the unified cell-type annotations, and the validation of MuCIA.

Comments on the Quality of English Language

The manuscript is understandable, but some minor language corrections are needed. There are several awkward expressions, for example, “how tense the interaction is?”. I suggest a comprehensive revision of the English throughout the manuscript.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

After reading the authors' article, I would like to express my appreciation. The authors have done a great deal of valuable work, as this information can save time for many users. The platform could serve as a useful tool, particularly for those working with spatial transcriptomics who wish to avoid starting from scratch every time.
However, I have a few questions for the authors:
Is there a clear workflow for using the platform?
While the article describes the modules and their functions, it lacks a clear, step-by-step usage scenario. The glioblastoma case study provides a single example but does not offer a generalized workflow. A schematic diagram would be very helpful. Is such a guide available in the "Document" module? If so, it should be mentioned in the article.
A second point concerns the following:
In Section 4.3, the CellPhoneDB v4.0 database is explicitly specified for COMMOT. For SpatialDM, the text mentions "extracting human LR pairs" (with a minimum threshold of 3 cells) but does not specify exactly which LR-pair database is used or whether it is applied to mouse samples, which constitute a significant portion of the platform's data. For CellNEST, the ligand-receptor pair database is not specified at all; only training parameters are provided (5 runs, 500 epochs, fixed seed). Since the authors cite the parallel comparison of three methods as a key advantage of the platform, it is necessary to clarify:

Which specific ligand-receptor pair databases are used for CellNEST and SpatialDM?
Is the same LR database applied to all three methods, or does each method operate using its own set?

There are some errors and typos, particularly in the references section.

Reference 8 and Reference 27 refer to the same article. Reference [8] is cited in the introduction in the context of CCL5–CCR5 in the tumor microenvironment—but the cited paper is clearly the wrong one. It appears the intended reference was lost. Additionally, Reference 8 contains an extra initial ("Z.X.") and omits the article number (3995), whereas Reference 27 lists this information correctly. In Reference 40, the word order and capitalization are incorrect:

LYMPHOCYTES-T MEDIATE IMMUNOLOGICAL CONTROL OF C3 GENE EXPRESSION. Its should be 

T lymphocytes mediate immunologic control of C3 gene expression.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf