Single-cell hi-c map prediction method based on single-cell RNA expression data

Abstract

The present invention belongs to the fields of bioinformatics and computational biology, and specifically relates to a single-cell Hi-C map prediction method based on single-cell RNA expression data. The method comprises the step of: constructing a pseudo-distance between scRNA expression data features and scHi-C map features by constructing a unified low-dimensional space, and mapping single-cell RNA expression data into a chromatin conformation space to achieve the goal of predicting an Hi-C map from RNA data. The present method can fully utilize the correlation between RNA expression data and chromatin conformation data, improving the accuracy and stability of prediction models, reducing the cost and technical difficulty of data acquisition, and providing new tools and ideas for studying gene expression regulation, cell differentiation, and disease mechanisms.

Description

Single-cell Hi-C map prediction method based on single-cell RNA expression data TECHNICAL FIELD

[0001] The present application belongs to the field of bioinformatics and computational biology, and specifically relates to a single-cell Hi-C map prediction method based on single-cell RNA expression data. BACKGROUND

[0002] Chromatin conformation plays a key role in gene expression regulation, cell differentiation and disease occurrence. Traditional chromatin conformation capture techniques (such as Hi-C) provide whole-genome level chromatin three-dimensional structure information, but usually require a large number of cells, making it difficult to capture single-cell level heterogeneity. The development of single-cell technology enables researchers to explore chromatin three-dimensional organization at the single-cell level. Chromatin is a complex structure composed of DNA, histone and other chromatin proteins, and is the storage and expression unit of genetic information in the nucleus. Chromatin not only has a linear structure, but more importantly, it forms a complex three-dimensional conformation in the nucleus. This three-dimensional structure plays a crucial role in biological processes such as gene expression regulation, cell cycle regulation, and DNA repair. Scientists have developed various 3C-based improved techniques, including 4C, 5C and Hi-C. 4C technology can analyze the interaction of chromatin fragments with other fragments of the whole genome, 5C technology can detect the interaction between multiple chromatin fragments, and Hi-C technology can detect the interaction between all chromatin fragments in the whole genome. The advantage of Hi-C technology is to provide high-resolution chromatin interaction data in the whole genome range, but it requires a large number of cell samples and cannot capture single-cell level chromatin heterogeneity.

[0003] The breakthrough of single-cell technology enables scientists to study various omics data such as genome, transcriptome and proteome at the single-cell level. Single-cell RNA sequencing (scRNA-seq) can detect gene expression in each cell at the single-cell level, revealing the functional state and gene regulation mechanism of the cell. Through scRNA-seq technology, researchers can identify different cell types and subtypes, understand the heterogeneity and dynamic changes of cells. Single-cell Hi-C technology (scHi-C) can capture the three-dimensional structure of chromatin at the single-cell level, compared with traditional Hi-C, scHi-C technology can reveal single-cell level chromatin heterogeneity, but the experimental cost is high, the technology is complex, the data noise is large and the coverage is low, which limits its application in large-scale research.

[0004] Data integration methods play an important role in genomics and transcriptomics research, enabling the extraction of comprehensive information from multidimensional data and revealing the rules of complex biological systems. Existing multi-omics data integration methods include integration based on paired data and unpaired data. Integration based on paired data requires different omics data from the same batch of cells, such as Seurat v3 and LIGER, which use canonical correlation analysis (CCA) or non-negative matrix factorization (NMF) to achieve joint embedding of data. Integration based on unpaired data processes data from different experiments by constructing a common low-dimensional space to map different omics data into the same space, such as Cobolt and MultiVI using variational autoencoder (VAE) to learn multi-omics data and reduce all cells' different modalities to a unified low-dimensional representation. GLUE (graph-linked unified embedding) learns feature embedding from prior knowledge through graph variational autoencoder, and then reconstructs and links omics data based on cell embedding.

[0005] Although these data integration methods have made significant progress in handling multi-omics data, they still face challenges when applied to single-cell RNA and single-cell Hi-C data integration. The sparsity and noise of single-cell Hi-C data increase the difficulty of data integration, and the feature space of single-cell RNA data and single-cell Hi-C data is quite different. How to accurately represent the features of these two types of data in a unified low-dimensional space is a key problem that data integration methods need to solve.

[0006] To address these challenges, the present invention proposes a deep learning-based single-cell RNA and single-cell Hi-C data integration method that maps single-cell RNA expression data to chromatin conformation space in a unified low-dimensional space, achieving the goal of predicting Hi-C maps from RNA data. This method can fully utilize the correlation between RNA expression data and chromatin conformation data, improve the accuracy and stability of the prediction model, reduce the cost and technical difficulty of data acquisition, and provide new tools and ideas for studying gene expression regulation, cell differentiation, and disease mechanisms. SUMMARY

[0007] The present invention first provides a single-cell Hi-C map prediction method based on single-cell RNA expression data, which includes the step of constructing the pseudo-distance between scRNA expression data features and scHi-C map features.

[0008] In certain embodiments, the step of constructing pseudo-distances between scRNA expression data features and scHi-C map features comprises using multi-modal alignment as an iterative optimization process under the guidance of feature embedding encoded by a graph, thereby mapping scRNA data and scHi-C data into the same low-dimensional space.

[0009] In certain embodiments, the mapping comprises (1) data preprocessing; and (2) constructing a cell embedding space.

[0010] In certain embodiments, the (1) data preprocessing comprises:

[0011] (1.1) Data acquisition: Obtain single-cell RNA sequencing data (scRNA-seq) and single-cell Hi-C data (scHi-C) from public databases or experimental data;

[0012] (1.2) Data filtering: Perform quality control on raw data to remove low-quality cells and genes. For RNA data, filter out lowly expressed genes; for Hi-C data, remove fragments with low interaction frequency;

[0013] (1.3) Data normalization: Normalize RNA data and Hi-C data to eliminate technical noise and batch effects. Common methods include log normalization and z-score standardization;

[0014] (1.4) Dimensionality reduction: Use methods such as principal component analysis (PCA) and t-SNE to reduce the dimensionality of the data, reducing the dimensionality of the data to facilitate subsequent model training and prediction.

[0015] In certain embodiments, the (2) constructing a cell embedding space comprises:

[0016] (2.1) Graph structure representation: Construct RNA data and Hi-C data into graph structure, nodes of RNA data represent genes, edges represent co-expression relationship between genes; nodes of Hi-C data represent genomic regions, edges represent interaction between regions.

[0017] (2.2) Correlation graph construction: By calculating the similarity between RNA data and Hi-C data features, construct the correlation graph of RNA and Hi-C data. Similarity calculation can use cosine similarity or Euclidean distance.

[0018] (2.3) Multi-modal data integration: Use GLUE and other unsupervised integration algorithms to integrate scRNA and scHi-C data into a common low-dimensional space using correlation graphs and dimensionality reduction results.

[0019] The application also provides a method for integrating single-cell RNA expression data and single-cell Hi-C maps, comprising the following steps:

[0020] (1) Normalizing and reducing dimension of RNA expression data and Hi-C maps.

[0021] (2) Constructing pseudo-distance of graph structure representation to characterize the relationship between the two features.

[0022] (3) Unsupervisedly mapping the two to the same low-dimensional space.

[0023] The application also provides a local single-cell Hi-C map prediction algorithm, comprising the following steps:

[0024] (1) Training with one-to-one data sets based on RNA data and Hi-C data.

[0025] (2) Using a transformer model as an encoder and a decoder to learn the mapping relationship between expression data and maps.

[0026] (3) Converting input data into a two-dimensional matrix and dividing it into fixed-size blocks to enhance the performance of the attention mechanism.

[0027] The application also provides a pseudo-time series analysis, model attribution, and label transfer method, comprising the following steps:

[0028] (1) Observing the dynamic changes of chromatin conformation at the single-cell level;

[0029] (2) Predicting chromatin regulators;

[0030] (3) Observing the differences in chromatin conformation under disease conditions;

[0031] (4) Applying integral gradient algorithm to analyze the model to identify regulators affecting chromatin conformation of specified genomic regions.

[0032] The application also provides applications of the above prediction methods, which are a) to achieve the goal of predicting Hi-C maps from RNA data; or b) to study gene expression regulation; or c) to study cell differentiation, or d) to study disease mechanisms, which are for non-diagnostic or non-therapeutic purposes.

[0033] Optionally, the applications also include pseudo-time series analysis, model attribution, and label transfer in chromatin conformation dynamic changes, regulator prediction;

[0034] The application also provides a model having a functional module for implementing the above method.

[0035] In some embodiments, the model comprises an encoder and a decoder.

[0036] In some embodiments, the encoder is a transformer; and the decoder is an autoencoder.

[0037] The present application finally provides an application of the above-mentioned model, which is a) to achieve the objective of predicting Hi-C maps from RNA data; or b) to study gene expression regulation; or c) to study cell differentiation, or d) to study disease mechanisms, which is for non-diagnostic or non-therapeutic purposes.

[0038] Optionally, the application further includes pseudo-temporal analysis, model attribution, and label transfer in chromatin conformation dynamics, regulator prediction.

[0039] Compared with the prior art, the present application has at least the following beneficial effects:

[0040] The SEE framework is trained using a large amount of single-cell RNA (scRNA) expression data and a small amount of single-cell Hi-C (scHi-C) maps as input.

[0041] The present application proposes a method for predicting single-cell Hi-C maps based on single-cell RNA expression data, which has many beneficial effects. The model can predict the Hi-C map of a specific genomic region at the single-cell level based on RNA expression data. For downstream applications, the trained model can infer the dynamic changes of chromatin conformation and identify chromatin regulators that affect the chromatin conformation of specific genomic regions, and can explore the impact of dynamic changes of chromatin conformation on disease occurrence and development. By combining deep learning models and multi-omics data integration techniques, the present application not only achieves the goal of inferring chromatin three-dimensional structure from RNA data in technology, but also significantly reduces experimental costs, improves data analysis efficiency, and enhances data integration capabilities, providing new tools and ideas for genomics and transcriptomics research. The following are the specific beneficial effects brought about by the present application:

[0042] 1. Reduce data acquisition costs

[0043] The experimental cost of single-cell Hi-C technology is high, which limits its application in large-scale studies. Each single-cell Hi-C experiment not only requires complex experimental operations, but also requires high reagent and equipment costs. In contrast, single-cell RNA sequencing technology has developed relatively maturely and can generate rich transcriptome data at a lower cost in a large number of cell populations. The present application predicts single-cell Hi-C maps by using existing single-cell RNA data, avoiding actual Hi-C experiments, thereby significantly reducing the cost of data acquisition. Researchers can obtain important information about chromatin three-dimensional structure without increasing additional experimental costs by using the method of the present application.

[0044] 2. Improve data analysis efficiency

[0045] Traditional chromatin conformation capture experiment procedures are complex, time-consuming, and require a large amount of computing resources for data processing and analysis. With the method of the present invention, researchers can use existing single-cell RNA data to quickly predict the three-dimensional structure of chromatin, saving a lot of time and resources. After the deep learning model is trained, it can predict new data at a high speed, greatly improving the efficiency of data analysis. This advantage makes the present invention have significant application value in large-scale data processing and high-throughput analysis.

[0046] 3. Enhance data integration capability

[0047] In genomics and transcriptomics research, the integration analysis of multi-omics data can provide more comprehensive biological information. The present invention integrates single-cell RNA and Hi-C data in a unified low-dimensional space, enabling joint analysis of different omics data. Using graph neural networks and variational autoencoders, the feature relationships in RNA data and Hi-C data are fully explored and represented. This multi-omics integration method not only improves the accuracy and stability of data analysis, but also provides a new perspective for studying the complex relationship between gene expression and chromatin structure.

[0048] 4. Improve research precision and depth

[0049] The present invention uses a deep learning model to learn the complex relationship between gene expression and chromatin three-dimensional structure from single-cell RNA data, enabling the prediction of chromatin interaction frequency with high precision. The model captures long-range dependencies and complex feature patterns in RNA data to generate high-quality Hi-C maps, improving research precision. In addition, by integrating different types of single-cell data, the present invention can reveal the gene expression regulation mechanism, cell differentiation process, and chromatin structure changes under disease conditions, providing new tools and methods for life science research.

[0050] 5. Wide applicability with strong generalization ability

[0051] The method of the present invention has strong generalization ability and can be applied to different types and sources of single-cell data. Through experiments on multiple public data sets, the model of the present invention has shown good stability and accuracy under different data sets and experimental conditions. Whether in human, animal or plant cells, the method of the present invention can effectively predict the three-dimensional structure of chromatin, with wide applicability. This feature makes the present invention have important application prospects in basic research, clinical application and biotechnology development.

[0052] 6. Promote the development of personalized medicine and precision medicine

[0053] Chromatin three-dimensional structure plays an important role in gene expression regulation and disease development. Through the method of the present application, the spatial conformation changes of disease-related genes can be revealed at the single-cell level, and the molecular mechanisms and regulatory networks of diseases can be understood. This capability is of great significance for the development of personalized medicine and precision medicine. Researchers and clinicians can use the method of the present application to analyze single-cell RNA data of patients, predict their chromatin three-dimensional structure, and develop personalized treatment plans to improve the accuracy of disease diagnosis and treatment.

[0054] 7. Promote the comprehensive utilization of multi-omics data

[0055] The integrated analysis of multi-omics data is an important trend in modern biological research. Through the method of the present application, researchers can jointly analyze single-cell RNA data with other omics data (such as proteomics, metabolomics, etc.) to comprehensively understand the functional state and regulatory mechanism of cells. The method of the present application provides an effective tool for the comprehensive utilization of multi-omics data and promotes the development of multi-omics research.

[0056] 8. Provide new research means for basic research

[0057] The present application not only has significant advantages in application, but also provides new research means for basic research. By inferring chromatin three-dimensional structure from single-cell RNA data, researchers can deeply explore the dynamic changes of chromatin in gene expression regulation, cell differentiation and development process. Especially in the study of the relationship between chromatin structure and function, the method of the present application has important scientific research value in revealing the chromatin regulation mechanism.

[0058] In summary, the technical scheme of the present application has significant beneficial effects in reducing data acquisition cost, improving data analysis efficiency, enhancing data integration capability, and improving research precision and depth. In addition, the method of the present application has wide applicability and strong generalization ability, promotes the development of personalized medicine and precision medicine, promotes the comprehensive utilization of multi-omics data, and provides new research means for basic research. Through the method of the present application, researchers can better understand the three-dimensional structure of the genome and its role in gene expression regulation and cell function, further promoting the development of life science research. BRIEF DESCRIPTION OF DRAWINGS

[0059] Figure 1: SEE model workflow diagram;

[0060] Figure 2: Integrated method architecture diagram;

[0061] Figure 3: Data flow diagram of mapping scRNA and scHi-C data;

[0062] Figure 4: Model architecture diagram;

[0063] Figure 5: Loss function curve during model training, showing the optimization of the loss function during training;

[0064] Figure 6: Comparison of predicted results and actual Hi-C maps, showing the accuracy and stability of the model prediction results;

[0065] Figure 7: Joint analysis of scHi-C maps and scRNA expression data of MBP gene sites on the pseudo-time scale;

[0066] Figure 8: Differences in changes of various features of QKI, DOCK10, and PIP4K2A gene sites on the pseudo-time scale;

[0067] Figure 9: Differences in changes of various features of QKI, DOCK10, and PIP4K2A gene sites on the pseudo-time scale;

[0068] Figure 10: Chromatin condition factors affecting dynamic changes in interactions;

[0069] Figure 11: Correlation analysis of oscillation patterns of gene expression counts and oscillation patterns of chromatin conformation volumes;

[0070] Figure 12: Robust interactions found around the gene BIN1 in astrocyte subtypes. DETAILED DESCRIPTION

[0071] To make the technical problems, technical solutions, and advantages of the present application clearer, the following will be described in detail with reference to the drawings and specific embodiments.

[0072] List of abbreviations, English, and key term definitions: 1, SEE (a Single-cell method for predicting Chromatin Conformation based on gene Expression);

[0073] 2, scRNA: single-cell RNA;

[0074] 3, scHi-C: single-cell Hi-C;

[0075] Example 1

[0076] A deep learning-based prediction model that can infer single-cell Hi-C maps from single-cell RNA expression data, the overall process and architecture of the model are shown in Figure 1. The specific technical solution includes the following steps:

[0077] As shown in FIG. 2, the present application proposes a method for integrating single-cell RNA (scRNA) expression data and single-cell Hi-C (scHi-C) mapping cell embeddings. To model the cell states, the present application normalizes and scales the scRNA expression data and scHi-C mapping to obtain their respective low-dimensional space information. With prior biological knowledge, the present application constructs the pseudo-distance between the scRNA expression data features and the scHi-C mapping features in a graph representation. Specifically, the vertices are genes and chromatin regions (i.e., 10-kilobase (kb) bins), and the positive edges will connect a gene and its genomic location. Then, under the guidance of the feature embedding encoded by the graph, the scRNA data and the scHi-C data are mapped into the same low-dimensional space using multi-modal alignment as an iterative optimization process. The SEE method is built on one scRNA expression dataset and one scHi-C mapping dataset. The mapping process of scRNA and scHi-C data is shown in FIG. 3.

[0078] 1. Data preprocessing:

[0079] (1.1) Data acquisition: Obtain single-cell RNA sequencing data (scRNA-seq) and single-cell Hi-C data (scHi-C) from public databases or experimental data.

[0080] (1.2) Data filtering: Perform quality control on raw data to remove low-quality cells and genes. For RNA data, filter out lowly expressed genes; for Hi-C data, remove fragments with low interaction frequency.

[0081] (1.3) Data normalization: Normalize RNA data and Hi-C data to eliminate technical noise and batch effects. Common methods include log normalization and z-score standardization.

[0082] (1.4) Dimensionality reduction: Use methods such as principal component analysis (PCA) and t-SNE to reduce the dimensionality of the data to facilitate subsequent model training and prediction.

[0083] 2. Constructing cell embedding space:

[0084] (2.1) Graph structure representation: Construct RNA data and Hi-C data into graph structure, nodes of RNA data represent genes, edges represent co-expression relationship between genes; nodes of Hi-C data represent genomic regions, edges represent interaction between regions.

[0085] (2.2) Correlation graph construction: Construct the correlation graph of RNA and Hi-C data by calculating the similarity between the features of RNA data and Hi-C data. Similarity calculation can use cosine similarity or Euclidean distance.

[0086] (2.3) Multimodal data integration: Use the correlation graph and dimensionality reduction results to integrate scRNA and scHi-C data into a common low-dimensional space using unsupervised integration algorithms such as GLUE.

[0087] Example 2

[0088] Model training: Use transformer as the specific model of encoder and decoder to learn the dependency between features (gene expression) in low information density input data (scRNA expression data), as shown in Figure 4. In order to enhance the performance of attention mechanism in utilizing a large number of features, the input data is converted into a two-dimensional matrix and then divided into multiple fixed-size small blocks, each of which is encoded separately in the input layer of the model. Since the focus of model learning is on initial information extraction to obtain factors affecting chromatin conformation, a more complex encoder network is needed to extract information from input data. See Figure 4.

[0089] (3.1) Model architecture design: Design a deep learning model based on transformer and autoencoder. The model includes two parts, encoder and decoder, the encoder is responsible for extracting the features of RNA data, and the decoder is responsible for mapping the features to Hi-C map.

[0090] (3.2) Encoder: Use transformer as the encoder to capture long-range dependencies and complex feature patterns in RNA data. Transformer consists of multiple layers of self-attention mechanism, which can effectively handle high-dimensional and sparse data.

[0091] (3.3) Decoder: Use autoencoder as the decoder to map the encoded low-dimensional representation back to high-dimensional space and generate Hi-C map. The decoder consists of multiple layers of fully connected neural network, which can reconstruct the high-dimensional interaction frequency matrix.

[0092] (3.4) Loss function: Use focal loss function for model training to handle data imbalance problem. Focal loss function can better handle rare classes and improve model performance on sparse data.

[0093] (3.5) Training process: Use Adam optimizer to train the model, and through cross-validation and parameter tuning, improve the generalization ability of the model. The training data set is divided into training set, validation set and test set, which are used for model training, validation and evaluation respectively.

[0094] Example 3

[0095] Prediction and Validation:

[0096] (4.1) Model Prediction: Apply the trained model to new single-cell RNA data to predict the corresponding Hi-C maps. The input data is pre-processed RNA expression data, and the output is the predicted chromatin interaction frequency matrix.

[0097] (4.2) Result Validation: Validate the prediction accuracy and stability of the model by comparing with actual Hi-C data. Common evaluation metrics include Pearson correlation coefficient, mean square error (MSE), and adjusted Rand index (ARI).

[0098] (4.3) Model Performance Evaluation: Conduct experiments on multiple public datasets to evaluate the model's generalization ability under different datasets and experimental conditions. Through cross-validation and multiple experiments, ensure the model has good stability and accuracy. See Figures 5-6. Figure 5 compares the focal loss and mean-square error (MSE) through experimental comparison, which can observe that the focal loss retains the overall accuracy of the MSE loss and produces better training results in the relatively small cell subtypes (Astro and OPC). Figure 6 shows that the predicted Hi-C maps are similar to the original results at Slc1a2 and Slc1a3 gene sites.

[0099] Downstream Analysis

[0100] (4.4) Pseudo-time Dynamics Analysis of Gene Expression and Chromatin Conformation: First, obtain the pseudo-time label of the sample by performing pseudo-time trajectory analysis on single-cell RNA (scRNA) expression data. Then, reorder the scRNA expression data and the predicted single-cell Hi-C (scHi-C) maps according to the pseudo-time label.

[0101] (4.5) Calculate Chromatin Volume: Use the predicted Hi-C maps to calculate the three-dimensional volume of chromatin in each cell. The specific method can be based on the three-dimensional reconstruction of the chromatin interaction frequency matrix to calculate the occupied volume of chromatin in three-dimensional space.

[0102] (4.6) Quantify Chromatin Interaction Dynamics Score: Analyze the reordered Hi-C maps to calculate the dynamics score of chromatin interaction. The dynamics score reflects the changes in chromatin interaction at different pseudo-time points.

[0103] (4.7) Visualization of TAD rearrangement: Analyze and visualize the rearrangement of TADs along the pseudo-time trajectory, study the dynamic changes of TAD boundaries and internal interactions. See FIG. 7. Taking MBP as an example, the UMAP plot in FIG. 7a intuitively shows that in the ODC cell subtype, the cells corresponding to the scRNA data are displayed in an obvious arc line, and this arc line has a significant correlation with the pseudo-time calculated using Scanpy. Subsequently, the cells are reordered according to the pseudo-time, and the gene expression values of the cells are visualized in the new order, as shown in FIG. 7b, it can be observed that the gene expression shows a clear upward trend as the pseudo-time progresses, which indicates that the expression of MBP changes with the development of ODC.

[0104] (4.8) Quantification of oscillation of gene locus-level chromatin interaction: Analyze the predicted Hi-C map to quantify the oscillation of chromatin interaction of specific gene loci along the pseudo-time trajectory. Calculate the interaction frequency of each gene locus at different pseudo-time points, and analyze its oscillation pattern. See FIGS. 8-9. As shown in FIG. 8, in order to study the intrinsic dynamics of chromatin conformation, this paper measures the distribution of changes in chromatin interaction. Since the threshold for determining whether the interaction oscillates is sample-specific in the calculation of switch score, it is more suitable for evaluating the dynamic changes between different interactions in the same sample. However, due to this processing method, switch score is not conducive to evaluating the dynamic changes between different samples. As shown in FIG. 9, for the same gene, a higher vibration score can be observed in the characterization of cell subtypes, indicating that the interaction is more dynamic compared to other cell subtypes.

[0105] (4.9) Obtain important regulators of chromatin: Use the integral gradient algorithm to calculate the importance of each feature (chromatin regulator) on the input data RNA expression data corresponding to a single interaction on the local Hi-C map; then, traverse all the interactions on the local Hi-C map to perform similar calculations, and take the average of the results as the contribution of all features on the RNA expression data to this local Hi-C map. See FIG. 10. As shown in FIG. 10, FIG. 10a and FIG. 10c represent the changes in the three-dimensional structure of the PDGFRA (a) / MBP (c) site during the differentiation process from OPC to ODC. FIG. 10b and FIG. 10d represent different regulatory patterns for different cell subtypes for the same genomic region (PDGFRA (b) and MBP (d)). Each heatmap consists of 20 chromatin regulators that are significantly different at three positions of the target gene (promoter, gene body, and terminator), and the color on the heatmap represents the importance score of the chromatin regulator calculated by the IG algorithm.

[0106] (4.10) Quantify the dynamic changes in chromatin conformation coupled with gene expression: By integrating gene expression data and chromatin conformation data on pseudo-temporal trajectories, quantify the dynamic changes in chromatin conformation coupled with gene expression. Analyze the coordinated change patterns of gene expression and chromatin interactions. See FIG. 11. As shown in FIG. 11, by analyzing up to 15 differentially expressed genes (DEGs) in ODC cell subtypes, we found a positive correlation between the oscillation amplitude of chromatin volume and the oscillation amplitude of expression level.

[0107] (4.11) Disease-associated single nucleotide polymorphisms can be explained by the dynamic features of chromatin conformation: Disease-associated single nucleotide polymorphisms (SNPs) can be explained by the dynamic features of chromatin conformation. By combining predicted Hi-C maps with disease-associated genes or variant data, study the role of chromatin conformation in disease. See FIG. 12. The results are shown in FIG. 12. In Astro cell subtypes, rs10207628 has robust interactions with the expression regulatory region. Therefore, we infer that these interactions are not easily changed in the disease state, thereby ensuring that the expression of BIN1 produces stable changes when the expression level of TFEB is up-regulated.

[0108] The above is the preferred embodiment of the present application, it should be noted that for those skilled in the art, without departing from the principles of the present application, can make several improvements and refinements, these improvements and refinements should also be considered as the protection scope of the present application.

Claims

1. A single-cell Hi-C mapping prediction method based on single-cell RNA expression data, characterized in that, The method comprises a step of constructing pseudo-distances between scRNA expression data features and scHi-C map features.

2. The prediction method of claim 1, wherein, The step of constructing pseudo-distances between scRNA expression data features and scHi-C map features comprises using multi-modal alignment as an iterative optimization process under the guidance of feature embedding encoded by a graph, so as to map scRNA data and scHi-C data into the same low-dimensional space.

3. The prediction method of claim 2, wherein, The mapping comprises (1) data preprocessing; and (2) constructing a cell embedding space.

4. The prediction method of claim 3, wherein, The (1) data preprocessing comprises: (1.1) data acquisition: acquiring single-cell RNA sequencing data (scRNA-seq) and single-cell Hi-C data (scHi-C) from public databases or experimental data; (1.2) data filtering: quality control of raw data, removal of low-quality cells and genes; for RNA sequencing data, filter out lowly expressed genes; for Hi-C data, remove fragments with low interaction frequency; (1.3) data normalization: normalizing RNA sequencing data and Hi-C data to eliminate technical noise and batch effects; common methods include log normalization and z-score standardization; (1.4) dimensionality reduction processing: using principal component analysis (PCA) and t-SNE to reduce the dimensionality of the data to facilitate subsequent model training and prediction.

5. The prediction method of claim 3, wherein, The (2) constructing a cell embedding space comprises: (2.1) graph structure representation: constructing RNA data and Hi-C data into a graph structure, the nodes of RNA data represent genes, and the edges represent the co-expression relationship between genes; the nodes of Hi-C data represent genomic regions, and the edges represent the interaction between regions; (2.2) correlation graph construction: by calculating the similarity between RNA data and Hi-C data features, construct the correlation graph of RNA and Hi-C data; similarity calculation can use cosine similarity or Euclidean distance; (2.3) multi-modal data integration: using the correlation graph and dimensionality reduction results, use the GLUE unsupervised integration algorithm to integrate scRNA and scHi-C data into a common low-dimensional space.

6. Use of a method of prediction according to any one of claims 1 to 5, characterized in that, The application is a) to achieve the goal of predicting Hi-C maps from RNA data; or b) to study gene expression regulation; or c) to study cell differentiation, or d) to study disease mechanisms, the application is for non-diagnostic or non-therapeutic purposes.

7. A model characterized by The model has a functional module for implementing the prediction method of any one of claims 1-5.

8. The model of claim 7, wherein, The model comprises an encoder and a decoder.

9. The model of claim 8, wherein, The encoder is a transformer; the decoder is an autoencoder (autoencoder).

10. Use of a model according to any one of claims 7-9, characterized in that, The application is a) to achieve the goal of predicting Hi-C maps from RNA data; or b) to study gene expression regulation; or c) to study cell differentiation, or d) to study disease mechanisms, the application is for non-diagnostic or non-therapeutic purposes.

Images