StereoSiTE: a framework to spatially and quantitatively profile the cellular neighborhood organized iTME

Article activity feed

1. 

   GigaScience

   Nov 23, 2024

   With emerging of Spatial Transcriptomics (ST) technology, a powerful algorithmic framework to quantitatively evaluate the active cell-cell interactions in the bio-function associated iTME unit will pave the ways to understand the mechanism underlying tumor biology. This study provides the StereoSiTE incorporating open source bioinformatics tools with the self-developed algorithm, SCII, to dissect a cellular neighborhood (CN) organized iTME based on cellular compositions, and to accurately infer the functional cell-cell communications with quantitatively defined interaction intensity in ST data. We applied StereoSiTE to deeply decode ST data of the xenograft models receiving immunoagonist. Results demonstrated that the neutrophils dominated CN5 might attribute to iTME remodeling after treatment. To be noted, SCII analyzed the spatially …

   With emerging of Spatial Transcriptomics (ST) technology, a powerful algorithmic framework to quantitatively evaluate the active cell-cell interactions in the bio-function associated iTME unit will pave the ways to understand the mechanism underlying tumor biology. This study provides the StereoSiTE incorporating open source bioinformatics tools with the self-developed algorithm, SCII, to dissect a cellular neighborhood (CN) organized iTME based on cellular compositions, and to accurately infer the functional cell-cell communications with quantitatively defined interaction intensity in ST data. We applied StereoSiTE to deeply decode ST data of the xenograft models receiving immunoagonist. Results demonstrated that the neutrophils dominated CN5 might attribute to iTME remodeling after treatment. To be noted, SCII analyzed the spatially resolved interaction intensity inferring a neutrophil leading communication network which was proved to actively function by analysis of Transcriptional Factor Regulon and Protein-Protein Interaction. Altogether, StereoSiTE is a promising framework for ST data to spatially reveal tumoribiology mechanisms.Competing Interest StatementThe authors have declared no competing interest.

   Reviewer 3. Rosalba Giugno

   Authors introduce StereoSiTE, which integrates open-source bioinformatics tools with the self-developed algorithm SCII. The aim is to dissect a cellular neighborhood (CN) organized iTME based on cellular compositions and accurately infer functional cell-cell communications with quantitatively defined interaction intensity in ST data.

   The paper's objective is commendable, and the overall organization of the content, along with the obtained results, holds great promise. Nevertheless, certain aspects need to be addressed. The proposed approach's novelty is significantly anchored in the SCII software. However, the paper has notable drawbacks. It falls short in providing a theoretical and scientific comparison with other similar tools. Moreover, the comparison includes systems that do not incorporate spatial considerations, posing a limitation in assessing the method's uniqueness in a broader context.

   Give more details on which systems are you referring here: "To improve accuracy, we recommended using spatially resolved data at single cell resolution". Please provide your insights on the rationale for employing or abstaining from downstream analysis to comprehend the spatial distribution of gene expression in tissue, as https://doi.org/10.1093/gigascience/giac075 and https://doi.org/10.1038/s41467-023-36796-3. Additionally, consider discussing how this is associated with the prediction, validation of the functional enrichment or on step: Clustering bins into different cellular neighborhoods based on their cellular composition.

   Re-reviews The authors have solved my issues.

   Read the original source
2. 

   GigaScience

   Nov 23, 2024

   AbstractWith emerging of Spatial Transcriptomics (ST) technology, a powerful algorithmic framework to quantitatively evaluate the active cell-cell interactions in the bio-function associated iTME unit will pave the ways to understand the mechanism underlying tumor biology. This study provides the StereoSiTE incorporating open source bioinformatics tools with the self-developed algorithm, SCII, to dissect a cellular neighborhood (CN) organized iTME based on cellular compositions, and to accurately infer the functional cell-cell communications with quantitatively defined interaction intensity in ST data. We applied StereoSiTE to deeply decode ST data of the xenograft models receiving immunoagonist. Results demonstrated that the neutrophils dominated CN5 might attribute to iTME remodeling after treatment. To be noted, SCII analyzed the …

   AbstractWith emerging of Spatial Transcriptomics (ST) technology, a powerful algorithmic framework to quantitatively evaluate the active cell-cell interactions in the bio-function associated iTME unit will pave the ways to understand the mechanism underlying tumor biology. This study provides the StereoSiTE incorporating open source bioinformatics tools with the self-developed algorithm, SCII, to dissect a cellular neighborhood (CN) organized iTME based on cellular compositions, and to accurately infer the functional cell-cell communications with quantitatively defined interaction intensity in ST data. We applied StereoSiTE to deeply decode ST data of the xenograft models receiving immunoagonist. Results demonstrated that the neutrophils dominated CN5 might attribute to iTME remodeling after treatment. To be noted, SCII analyzed the spatially resolved interaction intensity inferring a neutrophil leading communication network which was proved to actively function by analysis of Transcriptional Factor Regulon and Protein-Protein Interaction. Altogether, StereoSiTE is a promising framework for ST data to spatially reveal tumoribiology mechanisms.Competing Interest Statement

   Reviewer 2. Chenfei Wang

   In this manuscript, Xin. et al. provided a framework called StereoSiTE that incorporated the established methodologies with their developed algorithm to defined cellular neighborhood (CN) organized immune tumor microenvironment (iTME) based on cellular compositions, and to dissected the spatial cell interaction intensity (SCII) in spatial transcriptomics (ST). StereoSiTE has the following improvements compared to existing methods. First, SCII detects cell-cell communication using both cell space nearest neighbor graph and targeted L-R expression. Second, SCII taken the interaction distance account for different interaction classification such as secreted signaling, ECM receptor and cell-cell contact. Finally, StereoSiTE could avoided to detected the false positive interactions caused by limited reachable interaction.

   Although the authors performed comprehensive works to demonstrate the potential applications of StereoSiTE. This reviewer has strong concerns about the potential novelty and effectiveness of StereoSiTE over existing methods. The CN results were not mindful of the spatial information, and the labeled cellular neighborhood (CN) may mislead users. Additionally, although the L-R pair could be categorized into three classifications based on interaction distance, the SCII only uses different radius to infer cell communication without employing a different strategy for predicting interactions in distinct L-R pairs. I have the following detailed comments.

   Comments：

   1. The authors fail to show the novelty and advantages of CN compared to other methods, such as DeepST, which integrates gene expression, spatial location and image information. The authors should provide the comparison with the several recent strategies that consider the effect of local niches including BANKSY, stLearn, Giott, and DeepST.
   2. The authors should compare SCII with additional methods such as CellPhoneDB v3 and Cellchat v2, demonstrating its superior performance.
   3. The method used for cell segmentation should offer more comprehensive information rather than solely citing "Li, M. et al. (2023)".
   4. Format of the paper. The alignment inconsistency within the manuscript—with some paragraphs centered and others justified—should be corrected for uniformity.
   5. The figures and manuscript containing 'Teff' and 'M2-like' cell types should provide a legend explaining the abbreviations for clarity.
   6. The font size of the labels in Figures 5E-F is insufficient for easy reading and should be enlarged. Re-review: In the response letter, the author emphasizes the novelties of the StereoSiTE framework and demonstrates how the StereoSiTE software was specifically designed to address the question of "how iTME responds and functions under stimulation" using stereo-seq data. The author highlights notable enhancements to the self-development algorithm, including CN and SCII. The CN algorithm focuses on evaluating the cell composition in iTME, while SCII is designed to infer the intensity of spatial cell interactions. These advancements have been incorporated into the updated version of the manuscript. Notably, the SCII component of the framework combines spatial information and expression patterns to infer that cell-cell communication can limit reachable interactions, thereby reducing false positive interactions. The authors have also employed distinct strategies to predict different types of L-R pairs with varying interaction distances, encompassing secreted signaling, ECM-receptor, and cell-cell contact. In the case of secreted type L-R pairs, SCII enables the specification of varying radius thresholds to infer spatial cell communication. However, it is recommended that the authors consider the exponential decay of expression values, particularly when the radius exceeds 100 μm.

   The response also outlines the authors' claim that CN exhibits good performance compared to other tissue domain division methods (BANKSY and Giotto HMRF). However, upon reviewing the performance comparison results, it becomes apparent that BANKSY outperforms the other methods, although the CN method shows nearly consistent performance with BANKSY on the benchmark dataset STARmap. To substantiate the preference for CN over BANKSY, the authors are encouraged to provide evidence of its user-friendly interface, shorter run time, or lower memory usage. Overall, the revisions and enhancements made to the StereoSiTE framework significantly improve its functionality and analytical capabilities. The StereoSiTE software holds great promise in providing invaluable insights and support for potential users and researchers in the field.

   Read the original source
3. 

   GigaScience

   Nov 23, 2024

   AbstractWith emerging of Spatial Transcriptomics (ST) technology, a powerful algorithmic framework to quantitatively evaluate the active cell-cell interactions in the bio-function associated iTME unit will pave the ways to understand the mechanism underlying tumor biology. This study provides the StereoSiTE incorporating open source bioinformatics tools with the self-developed algorithm, SCII, to dissect a cellular neighborhood (CN) organized iTME based on cellular compositions, and to accurately infer the functional cell-cell communications with quantitatively defined interaction intensity in ST data. We applied StereoSiTE to deeply decode ST data of the xenograft models receiving immunoagonist. Results demonstrated that the neutrophils dominated CN5 might attribute to iTME remodeling after treatment. To be noted, SCII analyzed the …

   AbstractWith emerging of Spatial Transcriptomics (ST) technology, a powerful algorithmic framework to quantitatively evaluate the active cell-cell interactions in the bio-function associated iTME unit will pave the ways to understand the mechanism underlying tumor biology. This study provides the StereoSiTE incorporating open source bioinformatics tools with the self-developed algorithm, SCII, to dissect a cellular neighborhood (CN) organized iTME based on cellular compositions, and to accurately infer the functional cell-cell communications with quantitatively defined interaction intensity in ST data. We applied StereoSiTE to deeply decode ST data of the xenograft models receiving immunoagonist. Results demonstrated that the neutrophils dominated CN5 might attribute to iTME remodeling after treatment. To be noted, SCII analyzed the spatially resolved interaction intensity inferring a neutrophil leading communication network which was proved to actively function by analysis of Transcriptional Factor Regulon and Protein-Protein Interaction. Altogether, StereoSiTE is a promising framework for ST data to spatially reveal tumoribiology mechanisms.

   This work has been published in GigaScience Journal under a CC-BY 4.0 license (https://doi.org/10.1093/gigascience/giae078), and published as part of our Spatial Omics Methods series. The peer-reviews are as follows.

   Reviewer 1. Lihong Peng

   In this manuscript, the authors developed a computational framework named StereoSiTE to spatially and quantitatively profile the cellular neighborhood organized iTME by incorporating open source bioinformatics tools with their self-proposed algorithm named SCII. This study is very meaningful. However, it remains several problems.

   Major comments:

   1. The authors incorporated several open sources bioinformatics tools. However, how to ensure their combination is the optimal to the spatially resolved cell-cell communication inference performance? For example, cell2location was used to deconvolute cellular composition and construct cellular neighborhood. Why to use cell2location for deconvoluting spatial transcriptomics data? why not use the newest deconvolution algorithms, for example, SpaDecon, Celloscope, POLARIS, GraphST, SPASCER, and EnDecon? No model can adapt to all data. The authors should first verify that cell2location is the best appropriate cell type annotation tool corresponding to iTME. If not, the subsequent analyses will be not appropriate.

   2. The authors claimed that they computed the decomposition losses of different combinations of the number of CN modules and CT modules. Which combinations? The authors should list them.

   3. When measuring spatial cell interaction intensity, the authors only simply summed up the ligand and receptor gene expression information of the sender and receiver cells. Why not consider existing classical intercellular communication intensity methods? The authors should compare other intercellular communication intensity measurement methods. Please refer to the following two cites: Cell-cell communication inference and analysis in the tumour microenvironments from single-cell transcriptomics: data resources and computational strategies, briefings in bioinformatics. CellDialog: A Computational Framework for Ligand-receptor-mediated Cell-cell Communication Analysis, IEEE Journal of Biomedical and Health Informatics. Deciphering ligand-receptor-mediated intercellular communication based on ensemble deep learning and the joint scoring strategy from single-cell transcriptomic data, Computers in Biology and Medicine.

   4. For protein-protein interaction analysis, the authors queried 628 significant up regulated genes in CN5 area of treatment samples from STRING. Can all obtained proteins be ligands or receptors? In addition, they labeled hub genes and key protein-protein interaction networks, what were these hub genes and key networks used for?

   5. Which ligand-receptor pairs could mediate intercellular communication within immune tumor microenvironment? Among these L-R pairs, which L-R pairs are known in existing databases and which L-R pairs are the predicted ones?

   6. "The enrichment analysis of individual CN showed that each CN had a dominant cell type with a spatial aggregation (Fig 2F), which was increasingly obvious than that in whole slide (Fig 2E)." What's a dominant cell type? How to define it?

   7. "To reduce the variance among open-sourced L-R databases, we unified L-R database in SCII by choosing L-R dataset in CellChatDB, which assigned each L-R with an interaction distance associated classification as secreted signaling, ECM receptor and cell-cell contact." How to unify L-R database? Did it allow for user-specified LR databases and/or add user-specified LR databases?

   8. In figure 3, how to confirm which L-R pairs mediate intercellular communication?

   9. StereoSiTE is composed of multiple modules, is it scalable? Can some of these modules (such as clustering and cell type annotation) be replaced with other more powerful modules?

   10. The authors claimed that "CellPhoneDB detected many false positive interactions". How to find these false positive LRIs? How to validate the LRIs be false positives? Please list the found false positive LRIs.

   11. In Figure 3, the authors should add comparison experiments between StereoSiTME and classical intercellular communication analysis tools.

   Minor comments:

   1. The text in subfigure A, B, and C in Supplementary Figure 2 is obscure. The authors should revise Supplementary Figure 2.
   2. In Section "Abstract", iTME should use full name when it first appears.
   3. Which cites of "13 Li, M. et al. (2023)." is in the reference list?

   Re-review:

   In the revised manuscript, the authors conducted lots of revisions. However, it still remains many problems to solve:

   1. The authors have compared the performance of Cell2location with other cell type identification methods, Celloscope[10], GraphST[11], and POLARIS[12] on on both STARmap and stereo-seq dataset of liver cancer. How about its performance on other unlabeled datasets? Please compare it with "STGNNks: Identifying cell types in spatial transcriptomics data based on graph neural network, denoising auto-encoder, and 𝑘-sums clustering".

   2. Cell-cell communication is usually mediated by LRIs. The construction of high-quality LRI databases is very important to cell-cell communication. The authors should introduce these LRI data resources and potential LRI prediction methods and cite them, for example, PMID: 37976192, 37364528, 38367445.

   3. In Figure 4B, 4C, 4D, and 4F, Figure 5A and 5B, Figure 6B and 6C, the fonts are too small. Please enlarge the fonts.

   4. The organization and structure of this manuscript must be carefully revised. For example, The structure in Discussion is obscure. In the first paragraph in this section, the authors have introduced their proposed method, next, they described it in details. But the third paragraph elucidated the reason why to develop this reason. In addition, "Figure 3 highlights that the analysis without distance threshold may lead to false positive results, and SCII showed more superior performance than other methods." why to Figure 3? Did not the other results support their conclusion? The final paragraph in Discussion introduced their method again. It HAS NO logic.

   5. Where is the conclusion of this manuscript?

   6. The authors should analyze the limitations of this work for further work in the future.

   7. English is VERY POOR. This manuscript must be carefully revised. For example,

   "prove that spatial proximity is a must to guarantee an effective investigation.", is a must to do?

   Re-re-review: The authors have solved my issues.

   Read the original source
4. Version published to 10.1093/gigascience/giae078

   Jan 1, 2024
5. 

   Version published to 10.1101/2022.12.31.522366v3 on bioRxiv

   Sep 15, 2023
6. 

   Version published to 10.1101/2022.12.31.522366v2 on bioRxiv

   May 21, 2023
7. 

   Version published to 10.1101/2022.12.31.522366v1 on bioRxiv

   Dec 31, 2022