Genetic risk cell three-dimensional spatial positioning method, device, equipment and storage medium

Abstract

The present application relates to the technical field of bioinformatics, and discloses a genetic risk cell three-dimensional space positioning method, device and equipment and storage medium, the method comprises the following steps: acquiring spatial transcriptome data sets of multiple continuous tissue sections; determining the spatial propagation expression information corresponding to each detection site; determining the three-dimensional homogeneous detection site set of the target detection site; determining the three-dimensional gene-specific score value of the target gene based on the three-dimensional homogeneous detection site set; acquiring genome-wide association analysis data; performing association analysis based on the three-dimensional gene-specific score value of the target gene and the genome-wide association analysis data to obtain genetic risk association statistical results; and determining the spatial region related to the genetic risk of diseases and / or complex traits in the three-dimensional tissue reconstructed by the multiple continuous tissue sections based on the genetic risk association statistical results. The present application realizes the accurate spatial positioning of complex trait or disease related cells in three-dimensional tissue.

Description

Technical Field

[0001] This invention relates to the field of bioinformatics technology, specifically to a method, apparatus, device, and storage medium for three-dimensional spatial localization of genetically risky cells. Background Technology

[0002] Complex traits and common diseases are often formed by the combined effects of a large number of genetic variations. Genome-wide association studies (GWAS) can identify genetic loci associated with traits or diseases across the entire genome, but they can only verify the statistical association between genetic variations and diseases. They cannot identify the cell types, tissue regions, and spatial microenvironments that play a role in genetic risk, making it difficult to support clinical applications such as refined analysis of complex disease mechanisms, discovery of disease targets, and precision diagnosis and treatment.

[0003] Current research typically combines GWAS with single-cell RNA sequencing to infer disease-related cell types. However, this technique requires dissociating tissue samples, which completely destroys the original spatial location of cells, intercellular relationships, tissue structure, and local microenvironment information. It cannot reveal the spatial distribution patterns of genetic risks in tissues and is not suitable for tissue samples with complex three-dimensional structures such as the brain, embryos, tumors, and the heart. It is difficult to truly reflect the spatial positioning of disease-related cells in real tissues.

[0004] Spatial transcriptomics technology can preserve the in situ spatial information of gene expression, making up for the spatial defects of single-cell sequencing. However, traditional analysis methods are mostly limited to a single two-dimensional slice, which cannot restore the complete three-dimensional structure of the tissue. This can easily lead to the omission of key pathological areas, resulting in problems such as incomplete cell localization, analysis bias, and reduced statistical power.

[0005] However, most current technologies combining spatial transcriptomics and GWAS are based on a single two-dimensional slice, which can only preliminarily identify two-dimensional local regions related to genetic risk on a single two-dimensional slice, making it difficult to achieve precise three-dimensional spatial localization of cells related to complex traits or diseases. Summary of the Invention

[0006] This invention provides a method, device, equipment, and storage medium for three-dimensional spatial localization of cells at genetic risk, in order to solve the problem that existing combined spatial transcriptomics and GWAS analysis techniques are difficult to achieve accurate three-dimensional spatial localization of cells related to complex traits or diseases.

[0007] In a first aspect, the present invention provides a method for three-dimensional spatial localization of genetic risk cells, the method comprising: Acquire spatial transcriptome datasets from multiple consecutive tissue slices; the spatial transcriptome dataset includes gene expression information corresponding to each detection site on each tissue slice. For each tissue slice, based on the location information of the detection sites and gene expression information on the tissue slice, the spatial propagation expression information corresponding to each detection site is determined; Based on gene expression information and spatial propagation expression information, a three-dimensional homogeneous set of detection sites for the target detection site is determined; the detection sites in the three-dimensional homogeneous set of detection sites and the target detection site simultaneously satisfy spatial consistency and transcriptomic consistency. Based on a three-dimensional homogeneous set of detection sites, the three-dimensional gene-specificity score of the target gene is determined; the target gene is any gene in the target detection sites. Obtain genome-wide association analysis (GWIA) data; GWIA data includes SNP genetic statistics associated with disease and / or complex traits; Association analysis was performed based on the three-dimensional gene-specificity score of the target gene and genome-wide association analysis data to obtain the genetic risk association statistics of the target detection site; Based on the statistical results of genetic risk association, spatial regions associated with the genetic risk of disease and / or complex traits in a three-dimensional tissue reconstructed from multiple consecutive tissue slices were identified.

[0008] This invention constructs a dual-embedding characterization by collecting spatial transcriptome data from multiple consecutive tissue slices, and then screens out a set of three-dimensional homogeneous detection sites that possess both three-dimensional spatial continuity and transcriptome expression consistency. Based on this set of three-dimensional homogeneous detection sites, stratified heritability association analysis is performed in conjunction with genome-wide association analysis data. While preserving the three-dimensional spatial structure information of consecutive tissue slices, it jointly models transcriptome expression characteristics, local spatial context, cross-slice tissue continuity, and GWAS genetic statistical signals, effectively achieving precise spatial localization of cells associated with complex traits or diseases in three-dimensional tissues.

[0009] In one optional implementation, for each tissue slice, based on the location information of the detection sites and gene expression information on the tissue slice, the spatial propagation expression information corresponding to each detection site is determined, including: For each tissue slice, the adjacency relationship of the sites is constructed based on the Euclidean distance between each detection site within the tissue slice; Feature aggregation is performed on gene expression information of adjacent detection sites to generate spatial propagation expression information containing spatial structural features corresponding to each detection site.

[0010] It can further embed local spatial structure information in addition to gene expression information, thereby improving the model's ability to represent spatial organization patterns.

[0011] In one optional implementation, a three-dimensional homogeneous set of detection sites for the target detection site is determined based on gene expression information and spatial propagation expression information, including: Based on gene expression information, determine cell embedding information; Based on gene expression information and spatial propagation expression information, spatial domain embedding information is determined; Based on spatial domain embedding information, the spatial similarity between the target detection site and other detection sites is determined; Based on spatial similarity, a set of candidate detection sites that are spatially consistent with the target detection site is determined; Based on cell embedding information, transcriptomic similarity between target detection sites and candidate detection sites is determined; candidate detection sites are other detection sites in the candidate detection site set besides the target detection site. Based on transcriptome similarity, the top N candidate detection sites in the target tissue slice and its adjacent tissue slices are determined, and a three-dimensional homogeneous detection site set of the target detection site is generated; the target tissue slice is the tissue slice corresponding to the target detection site; N is an integer and N≥1.

[0012] This implementation method overcomes the limitations of traditional two-dimensional spatial analysis. By combining the original gene expression information and spatial structure propagation information of the detection site, it screens out three-dimensional neighboring sites that highly match the target detection site and have a consistent biological microenvironment within a three-dimensional tissue space composed of multiple consecutive tissue slices, constructing a three-dimensional homogeneous set of detection sites. This ensures both the continuity of the tissue's three-dimensional spatial structure and the homogeneity of the transcriptome's biology, significantly improving the accuracy of subsequent spatially specific gene scoring and genetic risk localization.

[0013] In an optional implementation, when the spatial transcriptome dataset also includes cell annotation information corresponding to each detection site on each tissue slice, the method includes: Cell embedding information and spatial domain embedding information are input into the classification network for supervised optimization to determine the main cell type corresponding to each detection site.

[0014] By using labeled supervised learning to optimize cell and spatial domain embedding features, it is possible to accurately distinguish the cell types to which each detection site belongs, which is beneficial for achieving refined spatial localization of disease genetic risk at the cellular level.

[0015] In one optional implementation, a three-dimensional gene-specific score for the target gene is determined based on a three-dimensional homogeneous detection site set, including: Based on gene expression information at each detection site, the original ranking of all genes within each detection site is determined. For the target gene in the target detection site, retrieve the middle ranking of the target gene in the three-dimensional homogeneous detection site set of each adjacent tissue slice; the target detection site is any one of the detection sites. Based on the intermediate ranking and the transcriptome similarity corresponding to the detection site, the ranking-weighted average of the target gene is determined; The three-dimensional gene-specific score of the target gene is determined based on the ranking-weighted average.

[0016] In this embodiment, a three-dimensional gene-specific score is obtained by weighted ranking calculation of neighboring detection sites in a three-dimensional homogeneous detection site set. This can effectively characterize the true enrichment degree of genes in three-dimensional tissue structures and help to achieve accurate spatial localization of disease-related genes.

[0017] In one optional implementation, after determining the three-dimensional gene-specific score of the target gene, the method further includes: If the expression of the target gene in the three-dimensional local area is not higher than the preset threshold of global background expression, then the three-dimensional gene specificity score corresponding to the target gene is assigned to zero. If the expression percentage of the target gene in the three-dimensional homogeneous detection site set is not higher than the preset judgment corresponding to the global background percentage, then the three-dimensional gene specificity score value corresponding to the target gene is assigned to zero.

[0018] Introducing an outlier filtering mechanism can effectively improve the robustness of the scoring results.

[0019] In one optional implementation, association analysis is performed based on the three-dimensional gene-specificity score of the target gene and genome-wide association analysis data to obtain genetic risk association statistics for the target detection site, including: The three-dimensional gene-specific score of the target gene is mapped to SNP sites within a preset regulatory window near the transcriptional region of the target gene, generating a set of SNP annotations corresponding to the target detection sites. A hierarchical linkage disequilibrium scoring regression framework was adopted, with the baseline annotation model as a conditional covariate. Based on the SNP annotation set, the heritability enrichment of diseases and / or complex traits in the corresponding SNP set was evaluated, and the genetic risk association statistics of the target detection loci were obtained.

[0020] Using three-dimensional gene-specific scoring to perform spatial functional hierarchical annotation of SNP loci, and accurately decomposing the heritability enrichment level of local tissue regions through the S-LDSC model, can effectively improve the accuracy of three-dimensional spatial localization of disease loci.

[0021] In a second aspect, the present invention provides a three-dimensional spatial positioning device for genetic risk cells, the device comprising: The transcriptome data acquisition module is used to acquire spatial transcriptome datasets from multiple consecutive tissue slices; the spatial transcriptome dataset includes gene expression information corresponding to each detection site on each tissue slice. The spatial determination module is used to determine the spatial propagation expression information corresponding to each detection site for each tissue slice, based on the location information of the detection sites and gene expression information on the tissue slice. The homogenization determination module is used to determine the three-dimensional homogenized detection site set of the target detection site based on gene expression information and spatial propagation expression information; the detection sites in the three-dimensional homogenized detection site set and the target detection site simultaneously satisfy spatial consistency and transcriptome consistency. The scoring determination module is used to determine the three-dimensional gene-specific score of the target gene based on a three-dimensional homogeneous detection site set; the target gene is any gene in the target detection site. The GWAS determination module is used to acquire genome-wide association analysis (GWAS) data; GWAS data includes SNP genetic statistics associated with diseases and / or complex traits. The association analysis module is used to perform association analysis based on the three-dimensional gene-specificity score of the target gene and genome-wide association analysis data to obtain the genetic risk association statistics of the target detection site. The spatial localization module is used to identify spatial regions in a three-dimensional tissue reconstructed from multiple consecutive tissue slices that are associated with the genetic risk of disease and / or complex traits, based on the results of genetic risk association statistics.

[0022] Thirdly, the present invention provides an electronic device, comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the computer instructions to perform the genetic risk cell three-dimensional spatial localization method of the first aspect or any corresponding embodiment described above.

[0023] Fourthly, the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to perform the three-dimensional spatial localization method for genetic risk cells described in the first aspect or any corresponding embodiment thereof.

[0024] It should be noted that the three-dimensional spatial positioning device, electronic device, and computer-readable storage medium for genetic risk cells provided by this invention correspond to the aforementioned three-dimensional spatial positioning method for genetic risk cells. Therefore, regarding the beneficial effects of the three-dimensional spatial positioning device, electronic device, and computer-readable storage medium for genetic risk cells, please refer to the description of the corresponding beneficial effects of the three-dimensional spatial positioning method for genetic risk cells above, and will not be repeated here. Attached Figure Description

[0025] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0026] Figure 1 This is a flowchart illustrating a method for three-dimensional spatial localization of genetic risk cells according to an embodiment of the present invention. Figure 2 This is another flowchart illustrating the method for three-dimensional spatial localization of genetic risk cells according to an embodiment of the present invention; Figure 3 This is a schematic diagram of the gsMap spatial positioning analysis results according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the two-dimensional distribution results of gsMap according to an embodiment of the present invention; Figure 5 This is a schematic diagram of GSS distribution according to an embodiment of the present invention; Figure 6 This is a UMAP visualization diagram of cells and microenvironment according to an embodiment of the present invention; Figure 7 This is a Manhattan diagram according to an embodiment of the present invention; Figure 8 This is a schematic diagram of the Fortis analysis results according to an embodiment of the present invention; Figure 9 This is a schematic diagram illustrating the spatial localization analysis results of risk cells for atrial fibrillation according to an embodiment of the present invention; Figure 10 This is a structural block diagram of a three-dimensional spatial positioning device for genetic risk cells according to an embodiment of the present invention; Figure 11 This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0028] It is understood that before using the technical solutions disclosed in the various embodiments of the present invention, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in the present invention and their authorization should be obtained in accordance with relevant laws and regulations through appropriate means.

[0029] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0030] Currently, 3D reconstruction technology for continuous spatial transcriptome sections is maturing, enabling spatial registration and reconstruction of multiple consecutive tissue sections into 3D tissue expression atlases. However, existing 3D spatial transcriptome technologies are primarily used for tissue morphology visualization and gene expression visualization. A unified analysis scheme that simultaneously integrates GWAS genetic statistical signals, spatial context, cellular transcriptome characteristics, and cross-section tissue continuity is lacking, making it difficult to achieve precise spatial localization of cells associated with complex traits or diseases.

[0031] In view of this, according to an embodiment of the present invention, a method for three-dimensional spatial localization of genetic risk cells is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0032] This embodiment provides a method for three-dimensional spatial localization of genetically at-risk cells, which can be automatically executed by a server, GPU workstation, cloud computing platform, local analysis terminal, or interactive visualization analysis system. Figure 1 This is a flowchart of a method for three-dimensional spatial localization of genetic risk cells according to an embodiment of the present invention, such as... Figure 1 As shown, the process includes the following steps: Step S101: Obtain a spatial transcriptome dataset from multiple consecutive tissue slices. The spatial transcriptome dataset includes gene expression information corresponding to each detection site on each tissue slice.

[0033] Specifically, multiple consecutive tissue slices corresponding to a biological tissue sample can be obtained, and the corresponding spatial transcriptome can be collected to obtain a continuous spatial transcriptome and generate a spatial transcriptome dataset. These multiple consecutive tissue slices can be continuous slices taken along the tissue thickness direction, with sequentially adjacent spatial positions and interconnected tissue structures, which can collectively cover the complete three-dimensional spatial structure of the target tissue.

[0034] Spatial transcriptome datasets contain at least one gene expression information corresponding to each detection site on each tissue slice (see reference). Figure 2 The gene expression matrix is ​​shown in the image. The detection sites are in situ gene capture sites (spots) formed on the surface of tissue sections by the space transcriptome sequencing platform. Each detection site corresponds to a small local region of the tissue and can capture the transcriptional expression signals of multiple cells within that region, thus obtaining quantitative information on the expression of all detected genes. Specific gene expression information may include the expression count, expression abundance, or normalized expression level of each gene.

[0035] "Spot" represents a data point in the spatial transcriptome data, regardless of whether it corresponds to single-cell or multi-cell resolution. For each spot, the original gene expression information and the expression information after spatial propagation are used as inputs. First, the system screens for hypervariable genes, selecting 2000 hypervariable genes by default to ensure that the embedding representation has sufficient information. Since the analysis object includes multiple consecutive ST slices, genes with stable hypervariability across different slices are preferentially selected. Specifically, for each slice, the normalized dispersion of each gene is calculated, and the results from multiple slices are aggregated to obtain a global set of hypervariable genes. All subsequent embedding learning is based on this set of hypervariable genes.

[0036] Furthermore, the spatial transcriptome dataset can also synchronously store auxiliary data such as spatial coordinates, slice number, and boundary location information of each detection site, which can be used for subsequent neighborhood screening, three-dimensional spatial location association, etc.

[0037] In some optional implementations, the spatial transcriptome dataset may also include cell annotation information and regional pathological annotation information corresponding to each detection site, for optional supervised optimization and cell type correction, etc.

[0038] Spatial transcriptome datasets can be read directly from local storage or retrieved from the cloud via a database interface after preprocessing.

[0039] Step S102: For each tissue slice, based on the location information of the detection sites and gene expression information on the tissue slice, determine the spatial propagation expression information corresponding to each detection site.

[0040] Specifically, for each individual tissue slice, the local neighborhood range corresponding to each detection site is determined based on the spatial location information and gene expression information of each detection site within the tissue slice. This neighborhood range can be a local region of the slice centered on the detection site with a preset spatial distance as a threshold, containing several adjacent detection sites spatially close to the detection site. By mining the local spatial context features of each detection site, spatially propagated expression information that combines its own transcriptional expression characteristics with the spatial structural characteristics of the surrounding tissue is generated. Unlike the feature form of a single detection site that only contains its own independent gene expression data, this embodiment fully utilizes the local spatial continuity features of the tissue slice to associate the genes and spatial features of adjacent detection sites.

[0041] In some alternative implementations, step S102 includes: Step S1021: For each tissue slice, construct the site adjacency relationship based on the Euclidean distance between each detection site within the tissue slice.

[0042] Specifically, the two-dimensional spatial coordinates of all detection sites within a single tissue slice can be retrieved, and the Euclidean distance between any two detection sites can be calculated pairwise. By using a pre-set spatial distance threshold, two detection sites with an Euclidean distance less than the threshold are identified as spatially adjacent sites. This process is repeated for all detection sites within the slice to complete the construction of the adjacency relationships for all detection sites.

[0043] Step S1022: Feature aggregation is performed on the gene expression information of adjacent detection sites to generate spatial propagation expression information containing spatial structural features corresponding to each detection site.

[0044] Based on the established site adjacency relationships, gene expression information from all spatially adjacent detection sites is collected, with each target detection site (any one of all detection sites) as the core. Graph convolution propagation can be used to perform unified feature aggregation on the gene expression information of the target detection site itself and that of adjacent detection sites. This allows the local tissue spatial structure information and associated expression features carried by adjacent detection sites to be superimposed onto the original expression features of the target detection site, generating a spatially propagated expression matrix containing local spatial structure features (see [reference]). Figure 2 (The spatial adjacency matrix in the model). After aggregation, each detection site forms a novel expression representation that integrates its own transcriptional features with the structural features of its surrounding spatial context, i.e., spatially propagated expression information. This allows for the embedding of local spatial structural information beyond gene expression information, thereby improving the model's ability to represent spatial organization patterns.

[0045] Step S103: Based on gene expression information and spatial propagation expression information, determine the three-dimensional homogeneous detection site set of the target detection site. The detection sites in the three-dimensional homogeneous detection site set simultaneously satisfy spatial consistency and transcriptomic consistency with the target detection site.

[0046] In some alternative implementations, step S103 includes: Step S1031: Determine cell embedding information based on gene expression information. Determine spatial domain embedding information based on gene expression information and spatial propagation expression information.

[0047] Reference Figure 2 As shown in region a, two independent encoders are used to encode gene expression information and spatial propagation expression information to generate cell embedding information and spatial domain embedding information.

[0048] The encoder employs a multi-head Transformer structure, which captures the dependencies between genes through a self-attention mechanism and adaptively aggregates relevant expression signals in sparse, high-noise spatial transcriptome data.

[0049] To eliminate batch effects between different ST slices, in some optional implementations, batch embeddings are further constructed, and a variational autoencoder (VAE) framework is employed to learn a unified latent space representation. Specifically, by introducing latent variables, cell embeddings and spatial domain embeddings are mapped to a shared latent space, and the Evidence Lower Bound (ELBO) is used as the optimization objective. The encoder outputs latent distribution parameters, and differentiability is achieved through reparameterization techniques, thereby supporting gradient backpropagation training of the model.

[0050] In the decoding phase, this embodiment employs an MLP-based decoder to probabilistically reconstruct gene expression counts, assuming they follow a zero-inflated negative binomial distribution to accommodate the overdispersion and zero-inflated characteristics commonly found in spatial transcriptome data. The decoder outputs zero-inflated probability parameters and expression mean parameters, while the dispersion parameters are shared and learned during training to improve model stability. Simultaneously, the system normalizes the reconstruction results using the total expression level of the target site (spot) to improve expression reconstruction accuracy.

[0051] By using a VAE-based dual embedding framework, batch correction and unified representation learning between transcriptome slices in different spaces are achieved. At the same time, cell identity information and local spatial context information are jointly captured, thereby enabling the identification of spatially consistent and transcriptome-homogeneous cell populations in a cross-slice three-dimensional space.

[0052] Step S1032: Based on spatial domain embedding information, determine the spatial similarity between the target detection site and other detection sites; based on spatial similarity, determine a set of candidate detection sites that are spatially consistent with the target detection site.

[0053] Reference Figure 2 As shown in region b, for each target detection site, this embodiment first calculates its spatial similarity to other detection sites based on spatial domain embedding information. Specifically, cosine similarity can be used to measure the proximity of different detection sites in their potential spatial representations, and candidate detection sites with spatial similarity to the target detection site higher than a preset threshold (default threshold is 0.5) are selected to ensure that these candidate detection sites are in a consistent local spatial environment. This step can effectively remove detection sites with large differences in spatial structure, thereby avoiding erroneous matching across tissue regions.

[0054] Step S1033: Based on cell embedding information, determine the transcriptomic similarity between the target detection site and candidate detection sites; candidate detection sites are other detection sites in the candidate detection site set besides the target detection site; based on transcriptomic similarity, determine the top N candidate detection sites in the target tissue slice and its adjacent tissue slices, and generate a three-dimensional homogenized detection site set for the target detection site; the target tissue slice is the tissue slice corresponding to the target detection site; N is an integer and N≥1.

[0055] Based on the aforementioned spatially consistent set of candidate detection sites, transcriptomic similarity is further calculated using cell embedding information. Specifically, for each candidate detection site, the cosine similarity of its cell embedding with the target detection site is calculated, and the top N most transcriptomically similar detection sites are selected from both the target tissue slice and its adjacent tissue slices as homogenized detection sites. This dual screening mechanism ensures that the selected detection sites are spatially adjacent and highly consistent in transcriptomic expression patterns, thereby achieving stable cell state alignment across slices. Finally, the homogenized detection sites from all adjacent slices together constitute a three-dimensional homogenized detection site set for the target detection site.

[0056] This implementation method overcomes the limitations of traditional two-dimensional spatial analysis. By combining the original gene expression information and spatial structure propagation information of the detection site, it screens out three-dimensional neighboring sites that highly match the target detection site and have a consistent biological microenvironment within a three-dimensional tissue space composed of multiple consecutive tissue slices, constructing a three-dimensional homogeneous set of detection sites. This ensures both the continuity of the tissue's three-dimensional spatial structure and the homogeneity of the transcriptome's biology, significantly improving the accuracy of subsequent spatially specific gene scoring and genetic risk localization.

[0057] Step S104: Based on the three-dimensional homogeneous detection site set, determine the three-dimensional gene-specificity score of the target gene. The target gene is any gene in the target detection site.

[0058] In some alternative implementations, step S104 includes: Step S1041: Based on the gene expression information of the detection sites, determine the original ranking of all genes within each detection site.

[0059] Within each detection site, all genes are first sorted in descending order of expression level, with higher expression levels corresponding to higher rankings. Each gene within that site is then assigned a unique original ranking number.

[0060] Step S1042: For the target gene in the target detection site, retrieve the middle ranking of the target gene in the three-dimensional homogeneous detection site set of each adjacent tissue slice; the target detection site is any one of the detection sites.

[0061] The three-dimensional homogeneous detection site set output by step S103 is called. This set does not exist only in a single tissue slice where the target detection site is located, but is distributed on the target slice and adjacent tissue slices in multiple spatially continuous layers above and below. Each site in the set has been verified by dual embedding features and simultaneously meets the dual constraints of three-dimensional spatial consistency with the target detection site and transcriptome expression consistency. All sites in this set are collectively referred to as neighborhood detection sites in this embodiment.

[0062] Subsequently, a traversal operation is performed on the three-dimensional homogenized detection site set, entering each neighboring detection site in the set one by one, retrieving the original ranking of all genes within each detection site that has been pre-generated, and recording the expression ranking value of the current target gene within each neighboring detection site, i.e., the intermediate ranking.

[0063] Step S1043: Based on the intermediate ranking and the transcriptome similarity corresponding to the detection site, determine the ranking weighted average of the target gene.

[0064] For each gene in the target detection site, a weighted average of the target gene expression ranking is calculated across the homogeneous detection site set of adjacent slices. The weights are determined by the cell embedding cosine similarity, thereby increasing the contribution of detection sites more closely related to the target detection site. Higher similarity indicates more consistent cell type and transcriptional characteristics between the two detection sites, resulting in a larger weighting for that detection site in the calculation. Sites with significant differences are automatically weighted down to reduce noise interference.

[0065] The weighted average of the target gene in the homogeneous detection site set of each adjacent slice was calculated. Then, the weighted average of the ranking in all adjacent slices was further averaged and compared with the average expression ranking of the gene in all global detection sites to measure whether the gene has a significantly high expression feature in the neighborhood of the target detection site.

[0066] The following example illustrates this, assuming the target detection site is Spot0; step S103 selects four homogeneous detection sites: Spot1, Spot2, Spot3, and Spot4; the target gene is GeneX; The weight is the cosine similarity between the target detection site Spot0 and the cell embeddings of each homogenized detection site: Spot1 similarity = 0.4, Spot2 similarity = 0.3, Spot3 similarity = 0.2, Spot4 similarity = 0.1; First, determine the expression ranking (middle ranking) of the target gene GeneX within each of the four homogenization detection sites: GeneX's ranking in Spot1: 3; GeneX's ranking in Spot2: 5; GeneX's ranking in Spot3: 6; GeneX's ranking in Spot4: 8; Calculate the weighted average of the rankings of individual tissue slice groups based on cosine similarity: Weighted value = R1×W1 + R2×W2 + R3×W3 + R4×W4; Substituting the data: =3×0.4+5×0.3+6×0.2+8×0.1=1.2+1.5+1.2+0.8=4.7; Since there is only one set of adjacent slices in this example, the overall value is still 4.7 (if there are multiple sets of multi-layer slices, just add the weighted results of each set and divide by the number of sets).

[0067] Assume that GeneX's global average ranking among all tens of thousands of detection sites in the tissue is 28; Construct a normalization formula (inverse normalization, mapping the range to 0-1): Normalized score = 1 - local ranking weighted average / global average ranking; Substituting: Normalized score = 1 - 4.7 / 28 = 0.8321.

[0068] Step S1044: Determine the three-dimensional gene-specific score of the target gene based on the ranking-weighted average.

[0069] To enhance the ability to distinguish between highly expressed genes, the normalization results can be further exponentially mapped to define the final three-dimensional gene-specific score (3D GSS value).

[0070] Because this method integrates spatial domain embedding and cellular embedding information, and jointly considers cross-slice spatial consistency and transcriptomic homogeneity, compared with traditional two-dimensional spatial specificity analysis methods, the method of this invention (gsMap3D) can more stably identify spatial functional genes in continuous tissues across slices, and effectively alleviate the over-smoothing problem and local abnormal noise problem in spatial transcriptome analysis. Finally, through the above-mentioned three-dimensional gene specificity scoring framework, spatial specificity gene identification and three-dimensional spatial expression pattern analysis across continuous tissue slices are achieved.

[0071] In this embodiment, a three-dimensional gene-specific score is obtained by weighted ranking calculation of neighboring detection sites in a three-dimensional homogeneous detection site set. This can effectively characterize the true enrichment degree of genes in three-dimensional tissue structures and help to achieve accurate spatial localization of disease-related genes.

[0072] Step S105: Obtain genome-wide association analysis (GWAS) data. GWAS data includes SNP genetic statistics associated with the disease and / or complex traits.

[0073] Specifically, genome-wide association analysis (GWAS) data includes genetic association statistics of a large number of single nucleotide variant (SNP) sites across the entire genome. These are statistical results of associations between SNP sites and complex traits such as diseases and height, obtained through large-scale population sequencing. The data mainly includes information such as SNP location, association P-value, and genetic effect.

[0074] Step S106: Based on the three-dimensional gene-specificity score of the target gene and the genome-wide association analysis data, perform association analysis to obtain the genetic risk association statistics of the target detection site.

[0075] Specifically, the three-dimensional gene-specific score can be used as the gene spatial expression weight to perform regional hierarchical annotation of SNP sites within the GWAS data. Heritability decomposition is then performed based on the S-LDSC hierarchical linear regression model to map SNP genetic variations to corresponding functional genes. Differential contribution weights are assigned based on the gene's specificity within the three-dimensional tissue microenvironment. Finally, the genetic enrichment of the target disease or complex trait in the three-dimensional local tissue region where the target detection site is located is quantified, and the corresponding significance P-value is output (refer to...). Figure 2 The genetic risk correlation statistics, such as the proportion of heritability shown in region c), enable the precise location and quantitative assessment of pathogenic genetic variations in the three-dimensional space of tissues.

[0076] Step S107: Based on the genetic risk association statistics, determine the spatial regions in the three-dimensional tissue reconstructed from multiple consecutive tissue slices that are associated with the genetic risk of disease and / or complex traits.

[0077] Based on the genetic risk association statistics of each target detection site, the genetic significance value of each target detection site can be mapped back to the two-dimensional coordinate system corresponding to the original multiple continuous tissue slices. Then, the overall three-dimensional tissue model is reconstructed according to the correspondence between the slice layers. The significance threshold is defined according to the risk indicators such as the P-value and the proportion of heritability contribution of each site. All detection sites with statistical results significantly higher than the threshold are screened out. The discrete high-risk sites are clustered and connected in three-dimensional space to delineate continuous and complete three-dimensional anatomical regions. Finally, the core susceptible spatial segments inside the three-dimensional tissue structure affected by the enrichment of genetic variations of diseases or complex traits are accurately located, and the three-dimensional visualization and localization of the disease risk area is completed intuitively.

[0078] This invention constructs a dual-embedding characterization by collecting spatial transcriptome data from multiple consecutive tissue slices, and then screens out a set of three-dimensional homogeneous detection sites that possess both three-dimensional spatial continuity and transcriptome expression consistency. Based on this set of three-dimensional homogeneous detection sites, stratified heritability association analysis is performed in conjunction with genome-wide association analysis data. While preserving the three-dimensional spatial structure information of consecutive tissue slices, it jointly models transcriptome expression characteristics, local spatial context, cross-slice tissue continuity, and GWAS genetic statistical signals, effectively achieving precise spatial localization of cells associated with complex traits or diseases in three-dimensional tissues.

[0079] This embodiment also provides a method for three-dimensional spatial localization of genetic risk cells, which can be automatically executed by a server, GPU workstation, cloud computing platform, local analysis terminal, or interactive visualization analysis system. The process includes the following steps: Step S201: Obtain a spatial transcriptome dataset from multiple consecutive tissue slices. The spatial transcriptome dataset includes gene expression information corresponding to each detection site on each tissue slice. For details, please refer to [link to relevant documentation]. Figure 1 Step S101 of the illustrated embodiment will not be described again here.

[0080] Step S202: For each tissue slice, based on the location information of the detection sites and gene expression information on the tissue slice, determine the spatial propagation expression information corresponding to each detection site. For details, please refer to [link to details]. Figure 1 Step S102 of the illustrated embodiment will not be described again here.

[0081] Step S203: Based on gene expression information and spatial propagation expression information, determine the three-dimensional homogeneous detection site set for the target detection site. The detection sites in the three-dimensional homogeneous detection site set simultaneously satisfy both spatial and transcriptomic consistency with the target detection site. For details, please refer to... Figure 1 Step S103 of the illustrated embodiment will not be described again here.

[0082] Step S204: Based on the three-dimensional homogeneous detection site set, determine the three-dimensional gene-specificity score of the target gene. The target gene is any gene within the target detection site. For details, please refer to [link to details]. Figure 1 Step S104 of the illustrated embodiment will not be described again here.

[0083] In some optional implementations, after determining the three-dimensional gene-specificity score of the target gene, the method further includes: if the expression of the target gene in the three-dimensional local area is not higher than a preset threshold of global background expression, then the three-dimensional gene-specificity score of the target gene is assigned to zero; if the expression ratio of the target gene in the three-dimensional homogeneous detection site set is not higher than a preset judgment corresponding to the global background ratio, then the three-dimensional gene-specificity score of the target gene is assigned to zero.

[0084] To improve the robustness of the scoring results, this embodiment also introduces an outlier filtering mechanism. When the three-dimensional local expression level of the target gene is not significantly higher than the global background expression, the corresponding 3D GSS is directly set to zero to avoid low-confidence enrichment signals.

[0085] Furthermore, the expression ratio of the target gene in the three-dimensional homogenized detection site set is compared with its global expression ratio; if the local ratio is not significantly higher than the global ratio (global background ratio), it is considered that the high expression may be driven by abnormal detection sites, and the corresponding score is also set to zero.

[0086] The above dual constraint mechanism can effectively reduce the impact of abnormal detection sites and noise expression on the results.

[0087] Step S205: Obtain genome-wide association analysis (GWAS) data. GWAS data includes SNP genetic statistics associated with the disease and / or complex traits. For details, please refer to [link to relevant documentation]. Figure 1 Step S105 of the illustrated embodiment will not be described again here.

[0088] Step S206: Based on the three-dimensional gene-specificity score of the target gene and the genome-wide association analysis data, perform association analysis to obtain the genetic risk association statistics of the target detection site.

[0089] In some alternative implementations, step S206 includes: Step S2061: Map the three-dimensional gene-specific score of the target gene to SNP sites within a preset regulatory window near the transcriptional region of the target gene, and generate an SNP annotation set corresponding to the target detection site.

[0090] After obtaining the three-dimensional gene-specific scores, a three-dimensional association analysis framework between spatial transcriptomics and GWAS traits was further constructed to identify the association between different spatial regions, cell types, and local tissue structures and the genetic risk of complex diseases.

[0091] This framework first establishes a mapping relationship between 3D GSS and genetic variation at the detection site level. Specifically, for each gene, the system maps its 3D GSS to a SNP site located near the transcription region, considering genetic variation within ±50kb upstream and downstream of the gene by default, thus forming a local regulatory window with a total length of 100kb. Furthermore, for SNPs outside the window but with long-range regulatory effects, this embodiment can further integrate SNP-gene regulatory relationships from external epigenetic maps, thereby including long-distance regulatory SNPs in the analysis.

[0092] Based on the above mapping relationship, the system can generate a corresponding SNP annotation set for each detection site, that is, use the 3D GSS value of the detection site to perform weighted annotation of its associated SNPs.

[0093] Step S2062: Using a hierarchical linkage disequilibrium scoring regression framework, with the baseline annotation model as a conditional covariate, the heritability enrichment of diseases and / or complex traits in the corresponding SNP annotation set is evaluated based on the SNP annotation set, and the genetic risk association statistics of the target detection site are obtained.

[0094] Each detection locus is treated as a SNP annotation set, and the hierarchical linkage disequilibrium score regression (S-LDSC) framework is used to assess whether SNPs with high 3D GSS values ​​are enriched for the heritability of specific complex traits or diseases. During computation, the system uses the baseline annotation model as a conditional covariate to reduce interference from existing functional annotations and improve the accuracy of spatially specific genetic signal identification. Since 3D spatial transcriptome data typically contain millions of detection loci, this embodiment further employs a JIT compilation GPU acceleration strategy to achieve efficient parallelization of the p-value calculation process in S-LDSC, thereby supporting ultra-large-scale 3D spatial genetic association analysis.

[0095] Based on the association analysis at the detection site level, this embodiment further constructs a genetic association statistical framework at the region level and cell type level to quantify the overall significance of the association between different tissue regions or cell populations and the target trait.

[0096] Specifically, the system uses the Cauchy combined test (CCT) to aggregate the p-values ​​of multiple detection sites within the same annotation category to obtain regional association statistics. Since the CCT method is highly sensitive to very small p-values, a small number of mislabeled or outlier detection sites may lead to an aberrant amplification of regional significance. Therefore, this embodiment further introduces a conservative outlier filtering strategy. Specifically, within each annotation category, the system first calculates the median and interquartile range of the negative logarithmic p-values ​​corresponding to all detection sites, and removes outlier detection sites exceeding the threshold of "median plus twice the interquartile range." Then, it re-aggregates the CCT using the remaining detection sites to obtain more robust regional association p-values. Furthermore, considering that while regional CCT results can be used to determine whether a region is related to the target trait, their stability is insufficient for directly comparing the association strength between different regions, the region ranking uses the median of the negative logarithmic p-values ​​corresponding to the top 5% of detection sites after filtering as the regional association strength indicator, thereby improving the stability and discriminative power of comparisons between different regions.

[0097] Using three-dimensional gene-specific scoring to perform spatial functional hierarchical annotation of SNP loci, and accurately decomposing the heritability enrichment level of local tissue regions through the S-LDSC model, can effectively improve the accuracy of three-dimensional spatial localization of disease loci.

[0098] Step S207: Based on the statistical results of genetic risk associations, identify the spatial regions in the three-dimensional tissue reconstructed from multiple consecutive tissue slices that are associated with the genetic risk of disease and / or complex traits. For details, please refer to [link to relevant documentation]. Figure 1 Step S107 of the illustrated embodiment will not be described again here.

[0099] All the above steps utilize GPU acceleration to improve the computational efficiency and scalability of large-scale spatial transcriptome data analysis. Upon completion of the analysis, the system automatically generates an interactive analysis report in HTML format and provides a 3D visualization of the results, including GSS distribution, cell-trait mapping results, and tissue spatial structure information. Furthermore, this embodiment also provides an online interactive visualization portal to support users in dynamically browsing, retrieving, and analyzing the 3D mapping results.

[0100] This invention, while preserving the three-dimensional spatial structure information of continuous tissue slices, combines transcriptome expression features, local spatial context, cross-slice tissue continuity, and GWAS genetic statistical signals to achieve precise spatial localization of cells with complex traits or disease-related characteristics in three-dimensional tissues.

[0101] In some optional implementations, when the spatial transcriptome dataset also includes cell annotation information corresponding to each detection site on each tissue slice, the method includes: Cell embedding information and spatial domain embedding information are input into the classification network for supervised optimization to determine the main cell type corresponding to each detection site.

[0102] When spatial transcriptome data includes cell annotation information, this embodiment can also introduce a classification decoder to supervise and optimize cell embeddings and spatial domain embeddings. Specifically, cell embedding information, spatial domain embedding information, and batch embeddings are jointly input into the classification network. By outputting cell category probabilities and training with a cross-entropy loss function, the ability of the embedding representation to distinguish cell types is further improved.

[0103] Since three-dimensional transcriptome data typically contains millions of detection sites, this embodiment can employ a downsampling training strategy to improve training efficiency. This strategy involves randomly sampling a portion of the detection sites for training while maintaining the proportion of different cell types, and then using the Adam optimizer to update the parameters.

[0104] By using labeled supervised learning to optimize cell and spatial domain embedding features, it is possible to accurately distinguish the cell types to which each detection site belongs, which is beneficial for achieving refined spatial localization of disease genetic risk at the cellular level.

[0105] In summary, to establish the association between three-dimensional gene expression patterns and trait-related genetic variations, a dual embedding representation of molecular and spatial features was first constructed across continuous spatial transcriptome slices. This process maps millions of cells to a unified shared latent space. The shared latent space includes two types of embedding representations: a cell embedding based on gene expression information to characterize cell identity, and a spatial domain embedding that combines gene expression information with spatial coordinates to characterize local tissue structure and spatial relationships. Based on this dual embedding representation, transcriptome similarity and local spatial context information across slice cells were further modeled to identify homogeneous cells. Subsequently, a three-dimensional gene specificity score was calculated. This score was obtained by aggregating the standardized gene expression ranking results in three-dimensional homogeneous cells and was used to quantify the high expression level and expression specificity of target genes in the local three-dimensional cellular environment. A higher three-dimensional gene specificity score indicates a higher expression level and stronger specificity of the corresponding gene in the target cell. After obtaining the three-dimensional gene-specific scores, a correlation analysis was conducted between the GWAS statistical results and the three-dimensional gene-specific scores using heritability stratification and enrichment analysis methods to quantify cell-trait associations. Specifically, stratified LDscore regression (S-LDSC) was used to assess the degree of genetic enrichment at the cellular level; simultaneously, for statistical associations at the tissue region level, Cauchy combination tests were used to aggregate the p-values ​​corresponding to multiple cells within a region, thereby obtaining the significance of associations at the region level.

[0106] The gsMap3D method proposed in this invention, compared to traditional two-dimensional spatial transcriptome analysis methods and existing single-slice spatial genetic association analysis frameworks, can achieve high-precision identification and analysis of cells and spatially functional genes related to complex traits within a three-dimensional space across slices. By constructing a dual-embedding unified representation learning framework based on variational autoencoders, it can simultaneously integrate cellular transcriptome features and local spatial structure information, achieving batch correction and unified latent spatial mapping between different spatial transcriptome slices, thereby significantly improving the stability and accuracy of cross-slice cell state alignment. Compared to methods based solely on gene expression or single spatial neighborhood modeling, this invention, through joint modeling of cell embedding and spatial domain embedding, can more accurately identify spatially consistent and transcriptome-homogeneous cell populations, and effectively reduce the error impact of over-smoothing and local noise in spatial transcriptome data. Furthermore, this invention proposes a three-dimensional gene-specific scoring calculation framework, which, through cross-slice homogeneous detection site aggregation and hierarchical statistical modeling, achieves robust identification of spatially functional genes in complex tissues, and further improves the robustness of the results through outlier filtering and proportional constraint mechanisms. Compared to traditional two-dimensional spatial specific analysis methods, this invention can recover continuous three-dimensional spatial expression patterns that are difficult to identify in single-slice analysis, thus more realistically reflecting the spatial molecular structural relationships within tissues. Furthermore, this invention establishes a three-dimensional association analysis framework between spatial transcriptomics and GWAS genetic signals, enabling the identification of spatial enrichment patterns of complex disease-related genetic effects at the detection site, region, and cell type levels. Large-scale spatial genetic association analysis is achieved through the S-LDSC and CCT statistical frameworks. By introducing a JIT compilation GPU acceleration strategy, this invention also supports efficient parallel computing at the scale of millions of detection sites, significantly improving the efficiency of ultra-large-scale three-dimensional spatial transcriptomics data analysis. Furthermore, this invention provides an interactive three-dimensional visualization and online analysis platform, enabling multi-slice joint rendering, three-dimensional GSS display, and dynamic exploration of trait-detection site associations, thereby improving the interpretability, transparency, and reproducibility of the analysis results. The gsMap3D disease risk cell spatial localization analysis results can be referenced... Figures 3 to 8 As shown. Among them, Figure 3 This is a schematic diagram of the gsMap spatial positioning analysis results; Figure 4 This is a schematic diagram of the two-dimensional distribution results of gsMap; Figure 5 A schematic diagram of GSS distribution (gene spatial score (gene-specific score value)); Figure 6This is a visualization diagram of UMAP (Uniform Manifold Approximation and Projection, a non-linear dimensionality reduction algorithm) for cells and microenvironments. The horizontal and vertical axes represent the two virtual dimensions UMAP_1 and UMAP_2 obtained by UMAP dimensionality reduction. The scale values ​​on the coordinates are only used to locate the position of the points. Different colors represent different tissue spatial annotation areas. Figure 7 For Manhattan map, Figure 8 This is a schematic diagram of the Fortis analysis results.

[0107] In practical applications, this invention further validates the effectiveness and biological significance of gsMap3D in elucidating the spatial genetic mechanisms of complex diseases. The study first applied gsMap3D to spatial transcriptome datasets from mouse embryos at two stages: E16.5 and E11.5. The E11.5 dataset contains 81 consecutive sagittal spatial transcriptome slices, covering approximately 6.6 million cells, providing high-resolution three-dimensional spatial tissue structure information. Compared to traditional two-dimensional spatial transcriptome analysis, which can only observe local expression patterns based on a single slice, this invention can recover the continuous three-dimensional spatial structure within tissues across multiple consecutive slices, thus providing a more complete analysis of the true spatial distribution patterns of cells related to complex diseases within tissues. Based on the above data, this invention successfully identified the continuous gradient distribution pattern of cardiac cells associated with atrial fibrillation (AF) in three-dimensional space. Specifically, in both the E16.5 and E11.5 embryonic stages, the cardiac region exhibited the most significant genetic association signal with AF. In E16.5 embryo data, the system observed a spatial gradient trend of AF-related signals in the cardiac region. However, due to the limited number of slices covering the cardiac region in this dataset, the spatial resolution was limited, making it difficult to fully characterize its continuous three-dimensional spatial structure. In the E11.5 high-density continuous slice data, gsMap3D further recovered a more continuous and high-resolution spatial organization pattern. The system found that AF-related signals in the cardiac region exhibited a spatial gradient that gradually increased from right to left, and this spatial distribution pattern was highly consistent with the three-dimensional anatomical structure during cardiac development, with the strongest correlation region mainly located in the developing left ventricle. This result demonstrates that the present invention can effectively recover continuous spatial genetic gradients in complex tissues, and such spatial organization patterns are difficult to identify using traditional single-slice analysis methods.

[0108] Furthermore, this invention also combined regional annotation analysis and differential expression analysis to verify the biological significance of the aforementioned spatial gradient. The results showed that highly expressed genes located in the left-sided region of the heart were significantly enriched in cardiac function-related biological processes such as muscle contraction, cardiomyocyte development, and ATP biosynthesis, indicating that the spatial gradient identified by this invention not only has statistical significance but also clear biological functional significance. In addition, the study further integrated human AF single-cell RNA sequencing data for cross-validation, revealing that marker genes associated with left-sided cardiac cells showed stronger differential expression levels in AF patients, thus further verifying the correlation between the identification results of this invention and the actual disease pathogenesis mechanism. These results demonstrate that this invention can not only stably restore continuous three-dimensional spatial organizational patterns in complex tissues but also reveal the fine distribution patterns of genetic risks of complex diseases in tissue space. (See gsMap3D for the results of spatial localization analysis of risk cells in atrial fibrillation.) Figure 9 As shown, -log10(P) represents significant enrichment, and the red area indicates significant enrichment of atrial fibrillation-related genes in cells at that spatial location.

[0109] In addition to analyses related to heart disease, this invention can also be used to analyze the spatial organization patterns of cells associated with different complex traits in developmental tissues. For example, during embryonic development, disease-related cells in different tissue regions can exhibit obvious regional clustering. By combining spatial domain embedding and cell embedding information, this invention can effectively identify these cell populations distributed continuously across sections and reveal their correspondence with tissue developmental structures. Because this invention employs a unified three-dimensional spatial modeling approach, it effectively avoids the spatial pattern loss problems caused by missing sections, local noise, and spatial fragmentation in traditional two-dimensional analysis, thereby improving the integrity and stability of complex tissue spatial structure reconstruction.

[0110] In summary, the above results demonstrate that the gsMap3D proposed in this invention not only enables unified representation learning and cross-slice cell alignment across different spatial transcriptome slices, but also stably identifies cells associated with complex traits and spatial functional genes within a three-dimensional space, further elucidating the spatial organizational patterns of disease genetic risk in tissues. Compared to traditional two-dimensional spatial transcriptome analysis methods, this invention exhibits significant advantages in spatial continuity restoration, identification of genetic associations in complex diseases, cross-slice tissue structure resolution, and computational efficiency for large-scale spatial data. It can be widely applied in fields such as complex disease mechanism research, embryonic development research, spatial pathological analysis, and precision medicine.

[0111] This embodiment also provides a three-dimensional spatial positioning device for genetic risk cells, which is used to implement the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0112] This embodiment provides a three-dimensional spatial positioning device for cells with genetic risk, such as... Figure 10 As shown, it includes: Transcriptome data acquisition module 301 is used to acquire spatial transcriptome datasets from multiple consecutive tissue slices; the spatial transcriptome dataset includes gene expression information corresponding to each detection site on each tissue slice. The spatial determination module 302 is used to determine the spatial propagation expression information corresponding to each detection site for each tissue slice based on the location information of the detection sites and gene expression information on the tissue slice. The homogenization determination module 303 is used to determine a three-dimensional homogenized detection site set of the target detection site based on gene expression information and spatial propagation expression information; the detection sites in the three-dimensional homogenized detection site set and the target detection site simultaneously satisfy spatial consistency and transcriptome consistency. The scoring determination module 304 is used to determine the three-dimensional gene-specific score of the target gene based on the three-dimensional homogeneous detection site set; the target gene is any gene in the target detection site. GWAS determination module 305 is used to acquire genome-wide association analysis (GWAS) data; GWAS data includes SNP genetic statistics related to disease and / or complex traits. The association analysis module 306 is used to perform association analysis based on the three-dimensional gene-specific score of the target gene and genome-wide association analysis data to obtain the genetic risk association statistics of the target detection site. Spatial localization module 307 is used to determine, based on genetic risk association statistics, spatial regions in a three-dimensional tissue reconstructed from multiple consecutive tissue slices that are associated with the genetic risk of disease and / or complex traits.

[0113] In one alternative implementation, the space determination module 302 includes: For each tissue slice, the adjacency relationship of the sites is constructed based on the Euclidean distance between each detection site within the tissue slice; Feature aggregation is performed on gene expression information of adjacent detection sites to generate spatial propagation expression information containing spatial structural features corresponding to each detection site.

[0114] In one optional implementation, the homogenization determination module 303 includes: Based on gene expression information, determine cell embedding information; Based on gene expression information and spatial propagation expression information, spatial domain embedding information is determined; Based on spatial domain embedding information, the spatial similarity between the target detection site and other detection sites is determined; Based on spatial similarity, a set of candidate detection sites that are spatially consistent with the target detection site is determined; Based on cell embedding information, transcriptomic similarity between target detection sites and candidate detection sites is determined; candidate detection sites are other detection sites in the candidate detection site set besides the target detection site. Based on transcriptome similarity, the top N candidate detection sites in the target tissue slice and its adjacent tissue slices are determined, and a three-dimensional homogeneous detection site set of the target detection site is generated; the target tissue slice is the tissue slice corresponding to the target detection site; N is an integer and N≥1.

[0115] In an optional implementation, when the spatial transcriptome dataset also includes cell annotation information corresponding to each detection site on each tissue slice, the device includes: The classification module is used to input cell embedding information and spatial domain embedding information into the classification network for supervised optimization, and to determine the main cell type corresponding to each detection site.

[0116] In one alternative implementation, the scoring determination module 304 includes: Based on gene expression information at each detection site, the original ranking of all genes within each detection site is determined. For the target gene in the target detection site, retrieve the middle ranking of the target gene in the three-dimensional homogeneous detection site set of each adjacent tissue slice; the target detection site is any one of the detection sites. Based on the intermediate ranking and the transcriptome similarity corresponding to the detection site, the ranking-weighted average of the target gene is determined; The three-dimensional gene-specific score of the target gene is determined based on the ranking-weighted average.

[0117] In one optional implementation, after determining the three-dimensional gene-specific score of the target gene, the method further includes: If the expression of the target gene in the three-dimensional local area is not higher than the preset threshold of global background expression, then the three-dimensional gene specificity score corresponding to the target gene is assigned to zero. If the expression percentage of the target gene in the three-dimensional homogeneous detection site set is not higher than the preset judgment corresponding to the global background percentage, then the three-dimensional gene specificity score value corresponding to the target gene is assigned to zero.

[0118] In one alternative implementation, the correlation analysis module 306 includes: The three-dimensional gene-specific score of the target gene is mapped to SNP sites within a preset regulatory window near the transcriptional region of the target gene, generating a set of SNP annotations corresponding to the target detection sites. A hierarchical linkage disequilibrium scoring regression framework was adopted, with the baseline annotation model as a conditional covariate. Based on the SNP annotation set, the heritability enrichment of diseases and / or complex traits in the corresponding SNP set was evaluated, and the genetic risk association statistics of the target detection loci were obtained.

[0119] The three-dimensional spatial localization device for genetic risk cells provided in this embodiment of the invention can execute the three-dimensional spatial localization method for genetic risk cells provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects for executing the method. Further functional descriptions of the above modules and units are the same as in the corresponding embodiments described above, and will not be repeated here.

[0120] Figure 11 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention.

[0121] The following is a detailed reference. Figure 11 This diagram illustrates a structural schematic suitable for implementing an electronic device according to embodiments of the present invention. The electronic device may include a processor (e.g., a central processing unit, graphics processor, etc.) 401, which can perform various appropriate actions and processes according to a program stored in read-only memory (ROM) 402 or a program loaded from memory 408 into random access memory (RAM) 403. The RAM 403 also stores various programs and data required for the operation of the electronic device. The processor 401, ROM 402, and RAM 403 are interconnected via a bus 404. An input / output (I / O) interface 405 is also connected to the bus 404.

[0122] Typically, the following devices can be connected to I / O interface 405: input devices 406 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 407 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; memory devices 408 including, for example, magnetic tapes, hard disks, etc.; and communication devices 409. Communication device 409 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 11 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown, and more or fewer devices may be implemented or have instead.

[0123] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 409, or installed from a memory 408, or installed from a ROM 402. When the computer program is executed by the processor 401, it performs the functions defined in the three-dimensional spatial localization method for genetic risk cells according to embodiments of the present invention.

[0124] Figure 11 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of the present invention.

[0125] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as recordable on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and subsequently stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the three-dimensional spatial localization method for genetic risk cells shown in the above embodiments is implemented.

[0126] A portion of this invention can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to the invention through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0127] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A method for three-dimensional spatial localization of cells with genetic risk, characterized in that, The method includes: A spatial transcriptome dataset of multiple consecutive tissue slices is obtained; the spatial transcriptome dataset includes gene expression information corresponding to each detection site on each tissue slice. For each tissue slice, based on the location information of the detection sites on the tissue slice and the gene expression information, the spatial propagation expression information corresponding to each detection site is determined; Based on the gene expression information and the spatial propagation expression information, a three-dimensional homogeneous set of detection sites for the target detection site is determined; the detection sites in the three-dimensional homogeneous set of detection sites and the target detection site simultaneously satisfy spatial consistency and transcriptomic consistency. Based on the three-dimensional homogeneous detection site set, a three-dimensional gene-specificity score of the target gene is determined; the target gene is any gene among the target detection sites. Acquire genome-wide association analysis (GWIA) data; the GWIA data includes SNP genetic statistics related to diseases and / or complex traits. Based on the three-dimensional gene-specificity score of the target gene and the genome-wide association analysis data, an association analysis was performed to obtain the genetic risk association statistics of the target detection site; Based on the genetic risk association statistics, spatial regions associated with the genetic risk of disease and / or complex traits are identified in the three-dimensional tissue reconstructed from the multiple consecutive tissue slices.

2. The method according to claim 1, characterized in that, For each tissue slice, based on the location information of the detection sites on the tissue slice and the gene expression information, the spatial propagation expression information corresponding to each detection site is determined, including: For each tissue slice, the adjacency relationship of the sites is constructed based on the Euclidean distance between each detection site within the tissue slice; Feature aggregation is performed on gene expression information of adjacent detection sites to generate spatial propagation expression information containing spatial structural features corresponding to each detection site.

3. The method according to claim 1, characterized in that, The determination of a three-dimensional homogeneous set of target detection sites based on the gene expression information and the spatial propagation expression information includes: Based on the gene expression information, cell embedding information is determined; Based on the gene expression information and the spatial propagation expression information, determine the spatial domain embedding information; Based on the spatial domain embedding information, the spatial similarity between the target detection site and other detection sites is determined; Based on the spatial similarity, a set of candidate detection sites that are spatially consistent with the target detection site is determined; Based on the cell embedding information, the transcriptomic similarity between the target detection site and candidate detection sites is determined; the candidate detection sites are other detection sites in the candidate detection site set besides the target detection site. Based on the transcriptome similarity, the top N candidate detection sites in the target tissue slice and its adjacent tissue slices are determined, and a three-dimensional homogeneous detection site set of the target detection site is generated; the target tissue slice is the tissue slice corresponding to the target detection site; N is an integer and N≥1.

4. The method according to claim 3, characterized in that, When the spatial transcriptome dataset also includes cell annotation information corresponding to each detection site on each tissue slice, the method includes: The cell embedding information and the spatial domain embedding information are input into a classification network for supervised optimization to determine the main cell type corresponding to each detection site.

5. The method according to claim 3, characterized in that, The determination of the three-dimensional gene-specific score of the target gene based on the three-dimensional homogeneous detection site set includes: Based on gene expression information at each detection site, the original ranking of all genes within each detection site is determined. For the target gene in the target detection site, retrieve the median ranking of the target gene in the three-dimensional homogeneous detection site set of each adjacent tissue slice; the target detection site is any one of all detection sites; Based on the intermediate ranking and the transcriptome similarity corresponding to the detection site, the ranking-weighted average of the target gene is determined; Based on the ranking-weighted average, the three-dimensional gene-specific score of the target gene is determined.

6. The method according to claim 5, characterized in that, After determining the three-dimensional gene-specific score of the target gene, the process also includes: If the expression of the target gene in the three-dimensional local area is not higher than the preset threshold of global background expression, then the three-dimensional gene specificity score corresponding to the target gene is assigned to zero; If the expression percentage of the target gene within the three-dimensional homogeneous detection site set is not higher than the preset judgment corresponding to the global background percentage, then the three-dimensional gene specificity score value corresponding to the target gene is assigned to zero.

7. The method according to claim 1, characterized in that, The association analysis is performed based on the three-dimensional gene-specific score of the target gene and the genome-wide association analysis data to obtain the genetic risk association statistics of the target detection site, including: The three-dimensional gene-specific score of the target gene is mapped to SNP sites within a preset regulatory window near the transcriptional region of the target gene to generate a set of SNP annotations corresponding to the target detection site. A hierarchical linkage disequilibrium scoring regression framework was adopted, with the baseline annotation model as a conditional covariate. Based on the SNP annotation set, the heritability enrichment of diseases and / or complex traits in the corresponding SNP set was evaluated, and the genetic risk association statistics of the target detection site were obtained.

8. A three-dimensional spatial positioning device for cells with genetic risk, characterized in that, The device includes: The transcriptome data acquisition module is used to acquire a spatial transcriptome dataset from multiple consecutive tissue slices; the spatial transcriptome dataset includes gene expression information corresponding to each detection site on each tissue slice. The spatial determination module is used to determine the spatial propagation expression information corresponding to each detection site for each tissue slice, based on the location information of the detection sites on the tissue slice and the gene expression information. A homogenization determination module is used to determine a three-dimensional homogenized detection site set of the target detection site based on the gene expression information and the spatial propagation expression information; the detection sites in the three-dimensional homogenized detection site set and the target detection site simultaneously satisfy spatial consistency and transcriptomic consistency. The scoring determination module is used to determine the three-dimensional gene-specific score of the target gene based on the three-dimensional homogeneous detection site set; the target gene is any gene in the target detection site. The GWAS determination module is used to acquire genome-wide association analysis (GWAS) data; the GWAS data includes SNP genetic statistics related to diseases and / or complex traits. The association analysis module is used to perform association analysis based on the three-dimensional gene-specific score of the target gene and the genome-wide association analysis data to obtain the genetic risk association statistics of the target detection site. The spatial localization module is used to determine, based on the genetic risk association statistics, the spatial regions in the three-dimensional tissue reconstructed from the multiple consecutive tissue slices that are associated with the genetic risk of disease and / or complex traits.

9. An electronic device, characterized in that, include: A memory and a processor are interconnected, the memory storing computer instructions, and the processor executing the computer instructions to perform the three-dimensional spatial localization method for genetic risk cells as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing a computer to perform the method for three-dimensional spatial localization of genetic risk cells as described in any one of claims 1 to 7.

Images