Intercellular signal transduction inference method and device based on optimal transmission of knowledge graph, equipment and medium

Abstract

The invention discloses an intercellular signal transduction inference method and device based on optimal transmission of a knowledge graph, equipment and a medium, and relates to the technical field of genetic engineering. The source gene expression data of different cells in the source cell type and the target gene table data of different cells in the target cell type are found according to the cell type database. On the basis of the ligand-receptor interaction knowledge graph, obtaining various pathways with an interaction relationship between the source cell type and the target cell type, and constructing a pathway set; and with the aim of minimizing the total transmission cost, determining an optimal transmission cost matrix of each path in the path set by adopting an scKGOT algorithm. According to the method, the intercellular channel can be more conveniently deduced, and the intercellular communication can be deduced by accurately simulating the intercellular ligand-receptor signal channel.

Description

Technical Field

[0001] The present application relates to the field of genetic engineering technology, and in particular to a method, device, equipment and medium for inferring intercellular signal transduction based on optimal transmission of knowledge graphs. Background Art

[0002] Single-cell RNA (Ribonucleic Acid) sequencing (scRNA-seq) technology has shown great potential in studying intercellular communication mediated by ligand-receptor interactions at the gene expression level. Currently, a variety of methods have been developed to directly predict active ligand-receptor pairs between two cell types based on gene expression and gene-to-gene correlation. However, due to the complex and intertwined nature of biological pathways, the activated signaling pathways that mediate intercellular communication cannot be reliably captured by simply examining the expression levels of ligand and receptor genes. To go a step further, tools such as NicheNet (NicheNet is an R package that calculates different cell-to-cell interactions) and CellCall (CellCall is a single-cell data intercellular communication analysis tool) are dedicated to identifying ligand-receptor pairs and their downstream genes, but existing methods are still limited in inferring intercellular communication using biological pathways, because ligand-receptor-mediated intercellular communication depends on the activation of specific signaling pathways. Therefore, accurately modeling the ligand-receptor signaling pathways between cells to infer intercellular communication remains a major challenge. Summary of the invention

[0003] The purpose of this application is to provide a method, device, equipment and medium for inferring intercellular signaling based on optimal transmission of knowledge graphs to accurately simulate the ligand-receptor signaling pathway between cells to infer intercellular communication.

[0004] To achieve the above objectives, this application provides the following solutions.

[0005] In a first aspect, the present application provides a method for inferring intercellular signal transduction based on optimal transmission of knowledge graphs, comprising: Obtain gene expression data of each cell of the source cell type from a cell type database as source gene expression data; Acquire gene expression data of each cell of the target cell type from a cell type database as target gene expression data; Based on the ligand-receptor interaction knowledge graph, each pathway that has an interaction relationship between the source cell type and the target cell type is obtained, and a pathway set is constructed; With the goal of minimizing the total transmission cost, the scKGOT (Single Cell Knowledge Graph Optimal transport) algorithm is used to determine the optimal transmission cost matrix for each pathway in the pathway set.

[0006] In a second aspect, the present application provides an intercellular signal transduction inference device based on optimal transmission of a knowledge graph, wherein the intercellular signal transduction inference device based on optimal transmission of a knowledge graph applies the above-mentioned intercellular signal transduction inference method based on optimal transmission of a knowledge graph, and the intercellular signal transduction inference device based on optimal transmission of a knowledge graph comprises: A source gene expression data acquisition module is used to acquire gene expression data of each cell of the source cell type from a cell type database as source gene expression data; A target gene expression data acquisition module is used to acquire gene expression data of each cell of the target cell type from a cell type database as target gene expression data; The candidate pathway acquisition module is used to obtain various pathways that have interaction relationships between source cell types and target cell types based on the ligand-receptor interaction knowledge graph and construct a pathway set; The path optimization module is used to determine the optimal transmission cost matrix of each path in the path set by using the scKGOT algorithm with the goal of minimizing the total transmission cost.

[0007] In a third aspect, the present application provides a computer device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-mentioned method for inferring intercellular signaling based on optimal transmission of knowledge graphs.

[0008] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned method for inferring intercellular signaling based on optimal transmission of knowledge graphs.

[0009] According to the specific embodiments provided in this application, this application has the following technical effects.

[0010] The present application provides a method, device, equipment and medium for inferring intercellular signaling based on optimal transmission of a knowledge graph. The present application finds source gene expression data of different cells in the source cell type and target gene table data of different cells in the target cell type according to the cell type database. According to the ligand-receptor interaction knowledge graph, the pathways are screened out and a pathway set is constructed. Then, with the goal of minimizing the total transmission cost, the scKGOT algorithm is used to determine the optimal transmission cost matrix for each pathway in the pathway set, which can more conveniently infer the pathways between cells and accurately simulate the ligand-receptor signaling pathways between cells to infer intercellular communication. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0012] Figure 1 A flowchart of a method for inferring intercellular signaling based on optimal transmission of a knowledge graph provided in one embodiment of the present application.

[0013] Figure 2 A schematic diagram of a method for inferring intercellular signaling based on optimal transmission of a knowledge graph provided in one embodiment of the present application.

[0014] Figure 3 This is a workflow diagram of the scKGOT algorithm provided in one embodiment of the present application.

[0015] Figure 4 An example diagram of a pathway provided for an embodiment of the present application.

[0016] Figure 5 This is a framework overview diagram of the scKGOT algorithm provided in one embodiment of the present application.

[0017] Figure 6 This is a performance comparison chart of the scKGOT algorithm provided in one embodiment of the present application and other algorithms.

[0018] Figure 7 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0019] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0020] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0021] Single-cell RNA sequencing (scRNA-seq) technology is increasingly used to characterize the heterogeneity of complex tissues. In addition to annotating cell types and transcript abundance, it is also very important to understand the underlying mechanisms of intercellular communication within the tissue microenvironment. Existing methods have difficulty fully utilizing biological pathways to infer intercellular communication because ligand-receptor-mediated intercellular communication relies on the activation of specific signaling pathways, such as the JAKSTAT (Janus Kinase Signal Transducer and Activator of Transcriptio, which is the main signaling mechanism of various cytokines and growth factors) pathway, PKC (Protein kinase C, PKC is a phosphatidyl alcohol signaling pathway with inositol-1,4,5-triphosphate and diacylglycerol as dual messengers) pathway, MAPK (Mitogen-Activated Protein Kinase, mitogen-activated protein kinase) pathway, etc. Accurately modeling intercellular ligand-receptor signaling pathways to infer intercellular communication remains a huge challenge. To this end, this application proposes the scKGOT algorithm, which relies on the knowledge graph optimal transmission algorithm to model and score the ligand-receptor-signaling network. scKGOT outperforms existing inference methods and demonstrates its superior accuracy and efficiency. In addition, some specific example studies are conducted to demonstrate its applicability in the real world. By combining scKGOT, we gain a deep understanding of the complex interactions of signaling pathways and the diverse cellular heterogeneity in specific biological contexts, greatly improving our understanding of the mechanisms of intercellular communication.

[0022] Inspired by the superior performance of optimal transfer in numerous tasks, we introduce here Single-cell Knowledge Graph Optimal Transfer (scKGOT), a method that constructs a signaling network (a union of multiple signaling pathways) from scratch to infer ligand-receptor-mediated intercellular communication for the first time using the proposed scKGOT algorithm based on single-cell transcriptome data. The method is first equipped with a Ligand-Receptor-Pathway Knowledge Graph (LRP-KG), which consists of intracellular and intercellular functional genes and gene interactions of different types of pathways. It then identifies activated signaling pathways by solving a novel scKGOT problem based on gene importance (i.e., the probability of functional interaction between two genes) and pathway performance (i.e., the Gromov-Wasserstein distance of the pathway), and identifies highly confident ligand-receptor pairs and top percentile pathways. Pathways involved in ligand-receptor-mediated intercellular communication are then reconstructed based on the predictions in the LRP-KG and known facts.

[0023] The proposed method for inferring intercellular signaling based on optimal knowledge graph transmission relies on the ligand-receptor interaction knowledge graph and makes full use of public single-cell transcriptome datasets. Comprehensive benchmark tests show that scKGOT outperforms existing inference methods in terms of accuracy and efficiency, demonstrating its excellent performance.

[0024] In an exemplary embodiment, Figure 1 and Figure 2 As shown, a method for inferring intercellular signal transduction based on optimal transmission of knowledge graphs is provided, including the following steps 101 to 104.

[0025] Step 101 : acquiring gene expression data of each cell of a source cell type from a cell type database as source gene expression data.

[0026] Step 102: Obtain gene expression data of each cell of the target cell type from a cell type database as target gene expression data.

[0027] Step 103, based on the ligand-receptor interaction knowledge graph, obtain each pathway having an interaction relationship between the source cell type and the target cell type, and construct a pathway set.

[0028] Step 104 , with the goal of minimizing the total transmission cost, the scKGOT algorithm is used to determine the optimal transmission cost matrix of each path in the path set.

[0029] Implementing the above steps 101 to 104 can more conveniently infer the pathway between cells and accurately simulate the ligand-receptor signaling pathway between cells to infer the cell-to-cell communication.

[0030] In another exemplary implementation, the following steps 105 and 106 are further provided after the above step 104 .

[0031] Step 105, according to the optimal transmission cost matrix of each pathway in the pathway set, the ligand-receptor gene score of each source gene and each target gene in the pathway set and the gene pathway score of each pathway in the pathway set are calculated.

[0032] Step 106 , using one or more of a heat map, a Circos map, a bubble map, a Sankey map, and a ridge map to visualize the optimal transmission cost matrix of each path in the path set.

[0033] like Figure 3 As shown in the figure, the above-mentioned intercellular signal transduction inference method based on optimal transmission of knowledge graph mainly includes three parts, such as Figure 3 a in Figure 3 b in Figure 3 As shown in c in the figure, the three parts are: data source, algorithm calculation process and cell-to-cell communication analysis diagram. The above-mentioned intercellular signaling inference method based on optimal transmission of knowledge graph aims to build a multi-pathway signal transduction network from scratch, focusing on ligand-receptor pairs and other related genes. The first part includes obtaining source gene expression data and target gene expression data from a cell type database. Exemplarily, in the embodiment of the present application, expression profiles and intercellular communication are extracted in the form of pathways in the cell type database-KEGG (Kyoto Encyclopedia of Genes and Genomes, Kyoto Encyclopedia of Genes and Genomes) database. The second part includes a biological model of intercellular communication and the calculation process of scKGOT. Given two cell types with scRNA-seq data, scKGOT formulates the intercellular gene-gene signal transmission process as a set of parallel optimal transmission models based on LRP-KG (i.e., a set of pathways). The method calculates the optimal transmission plan by minimizing the overall cost of transmitting gene expression in a shared space (i.e., the total transmission cost) to obtain the optimal transmission cost matrix. The plan includes binding scores for gene pairs (i.e., ligand-receptor gene scores) conditioned on biological pathways represented by LRP-KG subgraphs. The total cost of the process is quantified as the knowledge differences (KD) relative to each pathway. For the final prediction of ligand-receptor pairs, scKGOT integrates the best transmission plans from multiple biological pathways using a weighted strategy. The third part contains weighted transmission, signaling pathway discovery, and activation pathway knowledge graph analysis. To facilitate a comprehensive understanding of the results, this application provides a rich set of visualization tools, including heat maps, Circos plots, bubble plots, Sankey plots, and ridge plots.

[0034] In another exemplary implementation, the above-mentioned ligand-receptor interaction knowledge graph is constructed using the following steps 201 to 203.

[0035] Step 201 , collecting ligand-receptor interaction data based on data sets from different sources.

[0036] Step 202 , preprocessing the ligand-receptor interaction data to obtain cell expression data, cell type data, pathway data and ligand-receptor gene pair data.

[0037] Step 203, constructing the ligand-receptor interaction knowledge graph based on the cell expression data, cell type data, pathway data and ligand-receptor gene pair data.

[0038] In another exemplary implementation, in order to collect real ligand-receptor pairs, published papers related to scRNA-seq and intercellular communication were screened. Subsequently, the full text was carefully read and the potential ligand-receptors in the corresponding cell type pairs were focused. Their supporting data and related supplementary materials were accessed for verification, including the following datasets.

[0039] For the human liver bud scRNA-seq dataset (GSE96981), the true ligand-receptor pair (Endothelial Growth Factors A-Vascular Endothelial GrowthFactor Receptor 2, VEGFA-KDR, endothelial growth factor A-vascular endothelial cell growth factor receptor) was verified by inhibitor experiments in the original study; for the human embryo scRNA-seq dataset (GSE86146), the true ligand-receptor pair (DLL3-NOTCH2, NODAL-ACVR1C) was verified by immunostaining in the original study; for the human liver tumor scRNA-seq dataset (GSE146409), the true ligand-receptor pair (PDGFB-PDGFRB) was verified by smFISH imaging in the original study; for the human lung scRNA-seq dataset (GSE130148), the true ligand-receptor pair (HGF-MET) was verified by protein level co-localization experiments in another study. For the human placenta scRNA-seq dataset (E-MTAB-6701), the true ligand-receptor pairs (CSF1-CSF1R, CCL5-CCR1, CCL21-CCR7) were verified in other studies. For the two scRNA-seq datasets of human testis (scarface, testis), the true ligand-receptor pairs were verified by another study through immunostaining analysis.

[0040] For the mouse mammary gland scRNA-seq dataset (GSE148791), the true ligand-receptor pair (Tnf-Tnfrsf1a) was verified by inhibitor experiments in its original study. For the mouse spinal cord scRNA-seq dataset (GSE162610), the true ligand-receptor pair (Osm Osmr) was verified by immunohistochemistry analysis in its original study. For the mouse colon scRNA-seq dataset (SCP1038), the true ligand-receptor pair (Csr1-Csf1r) was found in its original study and verified by another study experiment. For the two scRNA-seq datasets of mouse brain (GSE129788, GSE165555), the true ligand-receptor pairs (Jag1-Notch1, Jag1-Notch2, Jag1-Notch3, Jag1-Notch4), Cx3cl1-Cx3cr1 were verified by other studies.

[0041] CellTalkDB (CellTalkDB is a comprehensive database of human and mouse ligand-receptor interaction pairs established by Professor Fan Xiaohui's research group at the School of Pharmacy of Zhejiang University through text mining and manual verification of known protein-protein interaction databases) provides a comprehensive collection of known ligand-receptor interactions, while KEGG is used to extract signaling pathway information.

[0042] In another exemplary embodiment, in the above step 202, four files are set to store cell expression data, cell type data, pathway data and ligand-receptor gene pair data respectively.

[0043] Among them, the cell expression data: the first row is the cell name, the first column is the gene name, which constitutes the expression levels of different genes in different cells and is a vector.

[0044] Cell type data: The second column is the cell name and the third column is the type.

[0045] Pathway data: The second column is the gene name, the third column is the gene name, the second and third columns represent the interaction between the source gene and the target gene. The fourth column is the pathway of gene interaction, the fifth column is the type of gene interaction, the sixth column is the direction of gene action, and the seventh column is the data source.

[0046] Ligand-receptor gene pair data: the second column is the ligand, receptor gene pair, the third column is the ligand gene, and the fourth column is the receptor gene.

[0047] In another exemplary embodiment, the method of screening pathways is: screening out all pathways whose gene coverage with the source gene and the target gene is greater than 0.3, and constructing a pathway set.

[0048] In another exemplary embodiment, the above-mentioned scKGOT algorithm utilizes LRP-KG as the backbone of signal transmission to computationally simulate the transmission of signal pathways. Currently, the optimal solution consists of two key parts: the optimal transmission cost matrix, which is used to identify ligand-receptor pairs, and the minimum cost of transmitting ligands to receptors through the predefined LRP-KG, which is called the knowledge difference (KD) of the pathway.

[0049] In another exemplary embodiment, the formula for calculating the total transmission cost is: ; in, For access The total transmission cost, is the correlation matrix between genes expressed in the source cell types, is the correlation matrix between genes expressed in the target cell type, For access Middle Gene expression data and The correlation value of the source gene expression data, For access Middle Target gene expression data and The correlation value of the source gene expression data, For access Middle The source gene is transported to the Characterization of the transmission cost of target genes, For access Middle The source gene is transported to the Characterization of the transmission cost of target genes.

[0050] like Figure 4 As shown, x is the source cell type, y is the target cell type, is the correlation matrix between genes expressed in the source cell types, is the correlation matrix between genes expressed in the target cell type. It is the transmission cost matrix of the path, that is, the transmission mode, which is the parameter that needs to be optimized and updated. express belong The joint distribution of ,pass To minimize the two distributions distance.

[0051] Next, the stable point is calculated step by step by iterating the Sinkhorn algorithm. In fact, the continuous linearization of the objective function (total transmission cost) leads to the continuity of the update, and the update is performed in a gradient descent manner. The formula is as follows: ; in, For the The path in the iteration process The transmission cost matrix is For the The path in the iteration process The transmission cost matrix is The path in the iteration process The gradient of the total transmission cost, is the correction factor.

[0052] In another exemplary embodiment, based on Second and The distance of the path in the Euclidean space during the iteration process is used to determine whether it has reached stability. The specific formula is as follows.

[0053] ; in, For the Second and The path in the iteration process t Distance in Euclidean space, Indicates The path in the iteration process Middle The source gene is transported to the Characterization of the transport costs of target genes, Indicates The path in the iteration process Middle The source gene is transported to the Characterization of the transport costs of target genes, represents the number of source genes, Indicates the number of target genes.

[0054] when Stop when it is less than 0.001 or the number of iterations is greater than 1000.

[0055] In another exemplary embodiment, in the above step 105, after the optimal transmission cost matrix is ​​obtained, the ligand-receptor gene scores of each source gene and each target gene in the pathway set and the gene pathway score of each pathway in the pathway set can be calculated based on the optimal transmission cost matrix of each pathway in the pathway set.

[0056] The ligand-receptor gene score was calculated as: ; in, For the The source gene and The ligand-receptor gene fraction of target genes, For access Middle The source gene is transported to the The optimal transmission cost representation of target genes, For access Middle The source gene is transported to the The optimal transmission cost representation of target genes, is the number of paths in the path set, It is a percentage ranking function, that is, the shorter the distance, the greater the weight.

[0057] The calculation formula for gene pathway score is: ; in, For access The gene pathway score of is the inverse normalization function, is the number of source genes, is the number of target genes. , a normalization method used to convert raw scores into inverse percentiles ranging from 0 to 1.

[0058] The ultimate goal is to use optimal transport to find pathways, signal transduction processes between cell types that connect ligands and receptors. This has greatly improved our understanding of the mechanisms of intercellular communication by providing a deep understanding of the complex interactions of signaling pathways and the diverse cellular heterogeneity in a specific biological context.

[0059] In another exemplary embodiment, in the above step 106, in addition to the existing visualization tools commonly used in the research field, the present embodiment also provides a comprehensive set of analysis tools specifically for intercellular signaling pathway analysis and joint ligand-receptor-pathway analysis. These tools include Circos diagrams for visualizing ligand and receptor activity, network visualizations depicting activated pathways and expressed genes, ridge diagrams containing additional pathway dimensions, and general interaction landscapes of ligands, receptors, and pathways involved in intercellular communication.

[0060] With the detailed predictions generated by scKGOT, researchers and biologists can intuitively explore and interpret the complex interactions between ligands, receptors, and signaling pathways, gaining valuable insights into cellular communication networks across species.

[0061] In order to illustrate the specific implementation of the above method, this application also provides the following examples.

[0062] To identify activated signaling pathways, scKGOT evaluates the importance of each gene within the pathway by calculating a gene importance score. It also evaluates pathway importance through the concept of pathway knowledge difference (KD), which measures the minimum cost of aligning the predicted gene interactions with the established structure of LRP-KG. By integrating these factors, scKGOT can pinpoint the most relevant pathways and identify the most critical genes in these pathways.

[0063] 1. Data construction.

[0064] scKGOT combines scRNA sequencing (scRNA-seq) data with a background knowledge graph called the Ligand-Receptor-Pathway Knowledge Graph (LRP-KG). LRP-KG is built by integrating detailed information from databases such as KEGG and CellTalkDB, providing a comprehensive resource to capture the complex relationships between ligands, receptors, and their associated signaling pathways.

[0065] CellTalkDB provides a comprehensive collection of known ligand-receptor interactions, while KEGG is used to extract signaling pathway information. The data is carefully curated to ensure high quality and biologically relevant interactions. Data processing involves the following steps.

[0066] Data retrieval: Ligand-receptor pairs and signaling pathway information were retrieved from CellTalkDB and KEGG, respectively.

[0067] Data Mapping: Genes and interactions were mapped to a common namespace to ensure consistency and compatibility between the two data sources.

[0068] Redundancy Removal: Duplicate entries and redundant interactions are identified and removed to avoid overrepresentation of certain gene pairs or pathways.

[0069] Ambiguity Resolution: Ambiguous gene names or symbols are resolved using standardized gene nomenclature and cross-references with other databases.

[0070] Biological Relevance Assessment: Curated interactions and pathways are assessed for biological relevance and potential impact on cell-to-cell communication based on literature evidence and expert knowledge.

[0071] All scRNA-seq datasets were retrieved from multiple high-quality reports. KEGG and Reactome provided 2,302 human pathways and 1,800 mouse pathways, with 2,223,641 and 1,651,421 ligand-receptor interaction data, including binding, dephosphorylation, and activation, respectively.

[0072] 2. scKGOT intercellular signal transmission process.

[0073] Incorporating an additional pathway dimension enables scKGOT to capture intercellular signaling processes at a more detailed level, thereby improving data separation and revealing complex relationships between gene pairs. This added dimension can better distinguish different clusters and accommodate complex patterns that may not be obvious in the original feature space, such as Figure 5 Once the activated signaling pathways are identified, scKGOT reconstructs these pathways by combining the model predictions with known information from LRPKG. This process involves integrating gene importance scores and pathway knowledge differences with existing biological data embedded in LRP-KG.

[0074] In order to illustrate the effect of the above method of the present application, the following test case is provided.

[0075] Experimental case 1: multi-level analysis of pathway activation.

[0076] scKGOT was used to conduct a comprehensive multi-level analysis of pathway activation from three perspectives: cells, genes, and pathways. Starting from the cellular level, various cell type pairs in the single-cell sequencing dataset were analyzed. Each pair of cell types was considered as the basic unit of study, and the method of this application was able to systematically explore the signaling pathways and interactions that are unique to each cell type pair, ensuring a thorough understanding of the cellular dynamics of the entire dataset. At the pathway level, observing the distribution of pathway knowledge difference scores revealed a significant deviation from the normal distribution, which supported the robustness of scKGOT predictions and showed that the method of this application effectively utilized pathway information without being biased by the size of the pathways involved. In particular, larger pathways did not have a disproportionate impact on the scores, which highlighted the size-insensitive nature of scKGOT in identifying and ranking pathways. In addition, through the internal mechanism of scKGOT, a detailed joint analysis of ligand-receptor-pathway interactions was performed. The heat map provides a multifaceted visualization that can capture the complexity of these interactions, thereby providing a deeper understanding of how scKGOT combines ligand-receptor interactions with pathway identification. This joint analysis is critical for understanding the underlying biological processes and validating the predictive power of the model.

[0077] Experimental case 2: scKGOT pathway interaction analysis.

[0078] LRP-KG was combined with optimal transport to analyze key pathways in different biological contexts. The effectiveness of scKGOT was demonstrated by its ability to identify the most important subgraphs within LRP-KG that showed high overlap with well-known pathways.

[0079] In an analysis of human liver buds, scKGOT identified the VEGFA-KDR axis as the center of the largest connected component in the LRP-KG subgraph. This axis was associated with the VEGF signaling pathway and was found to intersect with multiple key pathways, including PI3K / AKT, MAPK / ERK, and p38 MAPK. Some of these pathways spanned the entire subgraph, while others served as connectors between different signaling pathways, reinforcing the role of VEGFA-KDR as a key hub for information transmission within the liver bud microenvironment. The biological significance of this axis in angiogenesis and cell survival has been well documented. Similarly, in human embryos, scKGOT highlighted the DLL3-NOTCH2 axis as a key component of the largest connected subgraph, closely associated with the Notch signaling pathway. This pathway was found to be connected to other important pathways such as Wnt and Hedgehog, positioning DLL3-NOTCH2 as a central bridge of developmental signaling. In the human placenta, scKGOT identified the CSF1-CSF1R axis as an important element of the largest connected subgraph, which connected pathways such as MAPK / ERK, JAK / STAT, and NF-KB. These pathways are critical for immune regulation and trophoblast function, and CSF1-CSF1R is a key mediator of inter-pathway communication in the placental environment. Existing literature supports the importance of this axis in placental development and immune responses. The scKGOT algorithm not only successfully mapped the most relevant subgraphs in the LRP-KG to known biological pathways, but also revealed how these pathways interconnect to form key information bridges within cellular systems. The ability of scKGOT to identify these connections and overlaps demonstrates its effectiveness as a tool for exploring complex signaling networks. By focusing on the structural and functional relationships within the largest connected components of the LRPKG, scKGOT provides researchers with a powerful framework for studying multi-layer pathway interactions. This capability opens new avenues for understanding the complex mechanisms of cellular communication and signal transduction, providing a solid foundation for future biological research.

[0080] The method of the present application not only successfully mapped the most relevant subgraphs in LRP-KG to known biological pathways, but also revealed how these pathways interconnect to form key information bridges within cellular systems. The ability of scKGOT to identify these connections and overlaps demonstrates its effectiveness as a tool for exploring complex signaling networks. By focusing on the structural and functional relationships within the largest connected components of LRPKG, scKGOT provides researchers with a powerful framework for studying multi-layer pathway interactions. This capability opens new avenues for understanding the complex mechanisms of cellular communication and signal transduction, providing a solid foundation for future biological research.

[0081] Experimental case three: performance analysis in predicting ligand-receptor binding.

[0082] To evaluate the performance of our method in predicting ligand-receptor binding, we constructed this task as a multi-relation link prediction problem and compared scKGOT with several mature knowledge graph embedding (KGE) methods, including TransE, TransR, DistMult, RotatE, and ComplEx. These baseline methods consider various relational features, such as symmetric, asymmetric, reverse, combined, and 1-to-N relations. We conducted experiments on 6 human and 5 mouse scRNA-seq datasets, each containing at least one pair of ligands and receptors in a cell pair.

[0083] The evaluation protocol for ligand-receptor binding was designed as a multi-relation link prediction task, following the standard practice of link prediction under the stochastic local closed world assumption, with modifications to the overall ranking procedure. Due to the multiple pathways within the LRP-KG that a ligand-receptor pair may involve, all true triplets are evaluated and scores are assigned accordingly. The final ranking of each ligand-receptor prediction is determined by the highest rank achieved among all scored triplets, considering each path as a unique relation in the evaluation process. Different sizes of ranking candidates were considered in terms of reporting Mean Rank (MR), Mean Reciprocal Rank (MRR), and Hits@K (K=1, 10, 100, 500). scKGOT consistently outperforms these methods in terms of Hits@K and Mean Rank metrics, indicating that combining prior knowledge with gene expression data can significantly improve prediction accuracy.

[0084] To further validate the proposed scKGOT algorithm, we compared it with several baseline methods specifically designed for cell-cell interaction prediction, including CytoTalk, NicheNet, CellPhoneDB, SingleCellSignalR, CellChat, and CellCall. Figure 6 (a) is a comparison chart of the experimental results of 6 human scRNA-seq datasets. Figure 6 (b) is a comparison chart of the experimental results of 5 mouse scRNA-seq datasets. Figure 6 As shown in (a) and (b), the scKGOT algorithm achieves comparable accuracy levels to these baselines, and more importantly, scKGOT performs well in percentile rankings across all datasets, often positioning the target ligand-receptor pair at the top of the rankings, with values ​​frequently reaching 0.999. In addition, the boxplots show minimal variance, emphasizing the consistency of scKGOT in producing top predictions.

[0085] A noteworthy observation is that while traditional methods such as CytoTalk and NicheNet perform well in many cases, they can face challenges in scenarios involving modeling signal propagation through multiple pathways. In contrast, scKGOT's explicit consideration of pathway complexity appears to provide a more nuanced understanding of the complex dynamics of intercellular communication.

[0086] Overall, the scKGOT algorithm is robust in capturing structural and functional relationships between genes. The algorithm maintains consistently high accuracy in the top 10, 100, and 500 predictions, which is critical for practical applications where the accuracy of the top predictions is critical. These findings highlight the potential of scKGOT as a powerful tool for gene-level intercellular communication analysis, providing insights that may drive future biological discoveries.

[0087] Compared with the existing methods of inferring intercellular signals based on single-cell RNA sequencing (scRNA-seq) data, the intercellular signal transduction inference method based on optimal transmission of knowledge graphs in this application has three main advantages.

[0088] 1. LRP-KG, a knowledge graph specifically built for ligand-receptor-pathway interactions, is used to integrate biological prior knowledge to improve model accuracy.

[0089] 2. Use optimal transport to simulate the signal transduction process through pathways, connecting ligands and receptors.

[0090] 3. Provide comprehensive visualization functions, including heat maps and Sankey diagrams, to intuitively explore signaling pathways.

[0091] Based on the same inventive concept, the embodiment of the present application also provides a device for inferring intercellular signaling based on optimal transmission of knowledge graphs for realizing the above-mentioned method for inferring intercellular signaling based on optimal transmission of knowledge graphs. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above-mentioned method, so the specific limitations in one or more embodiments of the device for inferring intercellular signaling based on optimal transmission of knowledge graphs provided below can refer to the limitations of the method for inferring intercellular signaling based on optimal transmission of knowledge graphs above, and will not be repeated here.

[0092] In an exemplary embodiment, a cell-to-cell signaling inference device based on optimal transmission of knowledge graphs is provided, including the following modules.

[0093] The source gene expression data acquisition module is used to acquire the gene expression data of each cell of the source cell type from the cell type database as the source gene expression data.

[0094] The target gene expression data acquisition module is used to acquire the gene expression data of each cell of the target cell type from the cell type database as the target gene expression data.

[0095] The candidate pathway acquisition module is used to obtain various pathways with interaction relationships between source cell types and target cell types based on the ligand-receptor interaction knowledge graph and construct a pathway set.

[0096] The path optimization module is used to determine the optimal transmission cost matrix of each path in the path set by using the scKGOT algorithm with the goal of minimizing the total transmission cost.

[0097] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 7 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a cell-to-cell signal transduction inference method based on optimal transmission of a knowledge graph is implemented.

[0098] Those skilled in the art will understand that Figure 7 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components. In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0099] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0100] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0101] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0102] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0103] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0104] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for inferring intercellular signal transduction based on optimal transmission of knowledge graphs, characterized in that: include: Obtain gene expression data of each cell of the source cell type from a cell type database as source gene expression data; Acquire gene expression data of each cell of the target cell type from a cell type database as target gene expression data; Based on the ligand-receptor interaction knowledge graph, each pathway that has an interaction relationship between the source cell type and the target cell type is obtained, and a pathway set is constructed; With the goal of minimizing the total transmission cost, the scKGOT algorithm is used to determine the optimal transmission cost matrix of each path in the path set.

2. The method for inferring intercellular signal transduction based on optimal transmission of knowledge graph according to claim 1, characterized in that: The ligand-receptor interaction knowledge graph is constructed in the following way: Collect ligand-receptor interaction data based on datasets from different sources; Preprocess the ligand-receptor interaction data to obtain cell expression data, cell type data, pathway data, and ligand-receptor gene pair data; The ligand-receptor interaction knowledge graph is constructed based on cell expression data, cell type data, pathway data and ligand-receptor gene pair data.

3. The method for inferring intercellular signal transduction based on optimal transmission of knowledge graph according to claim 2, characterized in that: Datasets from different sources include: human liver bud scRNA-seq dataset, human embryo scRNA-seq dataset, human liver tumor scRNA-seq dataset, human lung scRNA-seq dataset, human placenta scRNA-seq dataset, mouse mammary gland scRNA-seq dataset, mouse spinal cord scRNA-seq dataset, and mouse brain scRNA-seq dataset.

4. The method for inferring intercellular signal transduction based on optimal transmission of knowledge graph according to claim 1, characterized in that: The formula for calculating the total transmission cost is: ； in, For access The total transmission cost, For access Middle Gene expression data and The correlation value of the source gene expression data, For access Middle Target gene expression data and The correlation value of the source gene expression data, For access Middle The source gene is transported to the j Characterization of the transmission cost of target genes, For access Middle The source gene is transported to the Characterization of the transmission cost of target genes.

5. The method for inferring intercellular signal transduction based on optimal transmission of knowledge graph according to claim 1, characterized in that: In the process of using the scKGOT algorithm to determine the optimal transmission cost matrix of each path in the path set, the update formula of the transmission cost matrix is: ； in, For the The path in the iteration process The transmission cost matrix is For the The path in the iteration process The transmission cost matrix is For the The path in the iteration process The gradient of the total transmission cost, is the correction factor.

6. The method for inferring intercellular signal transduction based on optimal transmission of knowledge graph according to claim 1, characterized in that: With the goal of minimizing the total transmission cost, the scKGOT algorithm is used to determine the optimal transmission cost matrix for each path in the path set, followed by: According to the optimal transmission cost matrix of each pathway in the pathway set, the ligand-receptor gene scores of each source gene and each target gene in the pathway set and the gene pathway score of each pathway in the pathway set are calculated; The ligand-receptor gene score was calculated as: ； in, For the The source gene and The ligand-receptor gene fraction of target genes, For access Middle The source gene is transported to the The optimal transmission cost representation of target genes, For access Middle The source gene is transported to the The optimal transmission cost representation of target genes, is the number of paths in the path set, is the percentage ranking function; The calculation formula for gene pathway score is: ； in, For access t The gene pathway score of is the inverse normalization function, is the number of source genes, is the number of target genes.

7. The method for inferring intercellular signal transduction based on optimal transmission of knowledge graph according to claim 1, characterized in that: With the goal of minimizing the total transmission cost, the scKGOT algorithm is used to determine the optimal transmission cost matrix for each path in the path set, followed by: One or more of a heat map, a Circos map, a bubble map, a Sankey map, and a ridge map are used to visualize the optimal transmission cost matrix of each pathway in the pathway set.

8. A cell-to-cell signaling inference device based on optimal transmission of knowledge graphs, characterized in that: The intercellular signal transduction inference device based on optimal transmission of knowledge graphs applies the intercellular signal transduction inference method based on optimal transmission of knowledge graphs according to any one of claims 1 to 7, and the intercellular signal transduction inference device based on optimal transmission of knowledge graphs comprises: A source gene expression data acquisition module is used to acquire gene expression data of each cell of the source cell type from a cell type database as source gene expression data; A target gene expression data acquisition module is used to acquire gene expression data of each cell of the target cell type from a cell type database as target gene expression data; The candidate pathway acquisition module is used to obtain various pathways that have interaction relationships between source cell types and target cell types based on the ligand-receptor interaction knowledge graph and construct a pathway set; The path optimization module is used to determine the optimal transmission cost matrix of each path in the path set by using the scKGOT algorithm with the goal of minimizing the total transmission cost.

9. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the intercellular signal transduction inference method based on optimal transmission of knowledge graphs as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, it implements the intercellular signal transduction inference method based on optimal transmission of knowledge graphs as described in any one of claims 1 to 7.

Images