Cell subpopulation classification method based on plant single cell transcriptome marker genes

Abstract

The application provides a cell subpopulation classification method based on plant single-cell transcriptome marker genes, an electronic device and a storage medium. Low-quality expressed original genes are removed based on a gene expression matrix to obtain a marker gene set, redundant information is effectively removed, key cell markers are accurately identified, and the classification accuracy of a subsequent trained model is improved. Then, a low-dimensional cell expression vector corresponding to each plant single cell is determined based on the marker gene set, a cell subpopulation classification model is obtained by training based on the low-dimensional cell expression vector corresponding to each plant single cell and the cell subpopulation category to which each plant single cell belongs, a low-dimensional cell expression vector of a plant cell to be classified is obtained based on the marker gene set, and the cell subpopulation to which the plant cell to be classified belongs is determined according to the low-dimensional cell expression vector of the plant cell to be classified and the cell subpopulation classification model, thereby realizing accurate classification of the plant cell to be classified.

Description

Technical Field

[0001] This invention belongs to the interdisciplinary field of plant single-cell bioinformatics and machine learning, and specifically relates to a cell subpopulation classification method, electronic device and storage medium based on plant single-cell transcriptome marker genes. Background Technology

[0002] Single-cell RNA sequencing (scRNA-seq) can resolve tissue composition and developmental trajectories at the single-cell scale, but accurately classifying cell subpopulations in plant scenarios still faces challenges: first, the expression matrix has high dimensionality and significant redundancy; second, differences between different tissues / species lead to insufficient model robustness and transferability; and third, the model needs to consider interpretability to support subsequent biological validation. Existing models are susceptible to noise and redundant genes on complex data, limiting their clustering and classification performance and resulting in low accuracy when classifying plant cells. Summary of the Invention

[0003] The purpose of this application is to provide a method, electronic device, and storage medium for classifying cell subpopulations based on plant single-cell transcriptome marker genes, in order to solve the above-mentioned technical problems.

[0004] On the one hand, a method for classifying cell subpopulations based on plant single-cell transcriptome marker genes is provided, including:

[0005] Acquire plant single-cell transcriptome sequencing data; the plant single-cell transcriptome sequencing data includes gene expression matrices corresponding to multiple plant single cells and cell subpopulation categories to which each plant single cell belongs, and each element value in the gene expression matrix represents the expression level of the corresponding original gene in the corresponding plant single cell;

[0006] Based on the gene expression matrix, low-quality expressed genes are removed to obtain a set of marker genes;

[0007] Based on the set of marker genes, a low-dimensional cell expression vector corresponding to each plant single cell is determined.

[0008] A cell subpopulation classification model is obtained by training based on the low-dimensional cell expression vector corresponding to each plant single cell and the cell subpopulation category to which each plant single cell belongs.

[0009] Based on the set of marker genes, obtain the low-dimensional cell expression vectors of the plant cells to be classified.

[0010] The cell subpopulation to which the plant cells to be classified belong is determined based on the low-dimensional cell expression vector of the plant cells to be classified and the cell subpopulation classification model.

[0011] In one embodiment, based on the gene expression matrix, low-quality expressed genes are removed to obtain a marker gene set, including:

[0012] Gene expression vectors corresponding to each of the original genes are obtained based on the gene expression matrix; the gene expression vector is a sequence composed of the expression levels of the original genes in each of the plant single cells;

[0013] For each of the original genes, the maximum information coefficient between the original gene and the category label is calculated based on the gene expression vector;

[0014] The marker gene set is obtained by removing low-quality expressed genes based on the maximum information coefficient.

[0015] In one embodiment, based on the maximum information coefficient, low-quality expressed original genes are removed to obtain a marker gene set, including:

[0016] The first candidate gene is obtained by removing the original genes whose maximum information coefficient is lower than the preset threshold;

[0017] For each first candidate gene, the gene expression vector corresponding to the first candidate gene is weighted based on the corresponding maximum information coefficient to obtain a weighted gene expression vector.

[0018] Based on the weighted gene expression vectors corresponding to each of the first candidate genes, the candidate gene expression matrix is ​​obtained;

[0019] Statistical analysis is performed on the element values ​​in the candidate gene expression matrix, and marker genes are determined from the first candidate genes based on the statistical analysis results; the marker gene set is composed of the marker genes.

[0020] In one embodiment, statistical analysis is performed on the element values ​​in the candidate gene expression matrix, and marker genes are determined from the first candidate genes based on the statistical analysis results, including:

[0021] Calculate the F-score value of each first candidate gene based on the candidate gene expression matrix;

[0022] The marker gene is determined from the first candidate gene based on the F-score value.

[0023] In one embodiment, determining the marker gene from the first candidate genes based on the F-score includes:

[0024] The first candidate genes are sorted in descending order of their F-score values;

[0025] Select the top M candidate genes by F-score from the first candidate genes after sorting;

[0026] The marker gene was determined from the second candidate gene based on cross-validation.

[0027] In one embodiment, a cell subpopulation classification model is obtained by training based on the cell subpopulation category to which each of the plant single cells belongs and the corresponding low-dimensional gene expression vectors, including:

[0028] The cell subpopulation category to which each plant single cell belongs and the corresponding low-dimensional gene expression vector are input into an initial model with a stacked ensemble structure for training to obtain a cell subpopulation classification model; the initial model includes multiple base models and a meta-model for fusing the initial prediction results of each base model to output the final prediction result.

[0029] In one embodiment, the training process includes:

[0030] In each round of cross-validation, each of the base models is trained on the first dataset of the training set, and the trained base models are used to predict the second dataset that was not trained, so as to obtain the prediction probability matrix of each base model for the second dataset.

[0031] For each round, the predicted probability matrices are stacked column-wise to form high-dimensional features;

[0032] The high-dimensional features and the corresponding cell subgroup categories to which each plant single cell belongs are used as inputs to the next layer meta-model. The meta-model learns and combines the prediction probability matrices of each base model to achieve dynamic weight allocation of the initial prediction results of each base model.

[0033] In one embodiment, the method is applied to an electronic device that stores a cell subpopulation classification model corresponding to each organ of each plant species;

[0034] Before obtaining the low-dimensional cellular expression vector of the plant cells to be classified based on the marker gene set, the method includes:

[0035] The corresponding cell subpopulation classification model is invoked based on the received prediction instruction, which includes the species and organ to which the plant cell to be classified belongs.

[0036] On the other hand, an electronic device is provided, including a processor and a memory, wherein the memory stores a computer program, and the processor executes the computer program to implement any of the methods described above.

[0037] On the other hand, a computer-readable storage medium is provided, characterized in that the computer-readable storage medium stores a computer program, which, when executed by at least one processor, implements any of the methods described above.

[0038] Beneficial Effects: The cell subpopulation classification method, electronic device, and storage medium based on plant single-cell transcriptome marker genes provided in this application acquire plant single-cell transcriptome sequencing data. This data includes gene expression matrices corresponding to multiple plant single cells and the cell subpopulation category to which each single cell belongs. Low-quality expressed genes are removed from the gene expression matrices to obtain a marker gene set, effectively eliminating redundant information and accurately identifying key cell markers, thus improving the classification accuracy of the subsequently trained model. Then, based on the marker gene set, low-dimensional cell expression vectors corresponding to each plant single cell are determined. A cell subpopulation classification model is obtained by training based on the low-dimensional cell expression vectors corresponding to each plant single cell and the cell subpopulation category to which each single cell belongs. Low-dimensional cell expression vectors of the plant cells to be classified are obtained based on the marker gene set. Finally, the cell subpopulation to which the plant cells to be classified belong is determined based on the low-dimensional cell expression vectors and the cell subpopulation classification model, achieving accurate classification of the plant cells to be classified.

[0039] Other features and advantages of this application will be set forth in the following description and will be apparent in part from the description or may be learned by practicing the application. The purposes and other advantages of this application may be realized and obtained by means of the structures particularly pointed out in the written description, claims, and drawings. It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not intended to limit this disclosure. Attached Figure Description

[0040] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0041] Figure 1 A flowchart illustrating the cell subpopulation classification method based on plant single-cell transcriptome marker genes provided in this application embodiment;

[0042] Figure 2 A flowchart illustrating the process of removing low-quality expressed genes from the gene expression matrix to obtain a marker gene set, as provided in this embodiment of the application.

[0043] Figure 3 A visual diagram illustrating the contribution of marker genes-based SHAP in experiments is provided for embodiments of this application.

[0044] Figure 4 This application provides a schematic diagram of UMAP based on marker genes for the experimental implementation of this application.

[0045] Figure 5 A schematic diagram of the ROC curves for the experiments provided in this application embodiment;

[0046] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0048] Single-cell RNA sequencing (scRNA-seq) technology can reveal the roles of various cells in life processes, but accurate classification of cell subpopulations still faces bottlenecks. Therefore, this application provides a method for classifying cell subpopulations based on plant single-cell transcriptome marker genes. Please refer to [link to relevant documentation]. Figure 1 As shown, it includes:

[0049] S11: Obtain plant single-cell transcriptome sequencing data.

[0050] The plant single-cell transcriptome sequencing data includes gene expression matrices corresponding to multiple plant single cells and the cell subpopulation category to which each plant single cell belongs. Each element value in the gene expression matrix represents the expression level of the corresponding original gene in the corresponding plant single cell.

[0051] S12: Based on the gene expression matrix, remove the original genes with low-quality expression to obtain the marker gene set.

[0052] S13: Determine the low-dimensional cell expression vector corresponding to each plant single cell based on the marker gene set.

[0053] S14: A cell subpopulation classification model is obtained by training based on the low-dimensional cell expression vector corresponding to each plant single cell and the cell subpopulation category to which each plant single cell belongs.

[0054] S15: Obtain low-dimensional cell expression vectors of plant cells to be classified based on the marker gene set.

[0055] S16: Determine the cell subpopulation to which the plant cells to be classified belong based on the low-dimensional cell expression vectors and cell subpopulation classification models of the plant cells to be classified.

[0056] The following is a detailed explanation of each of the above steps.

[0057] Please see Figure 2 As shown, step S12 may include the following sub-steps:

[0058] S121: Obtain the gene expression vector corresponding to each original gene based on the gene expression matrix; the gene expression vector is a sequence composed of the expression levels of the original gene in each plant single cell.

[0059] S122: For each original gene, calculate the maximum information coefficient between the original gene and the category label based on the gene expression vector.

[0060] S123: Based on the maximum information coefficient, remove the original genes with low-quality expression to obtain the marker gene set.

[0061] For example, step S123 may include the following:

[0062] Sub-step 1: Remove the original genes whose maximum information coefficient is lower than the preset threshold to obtain the first candidate gene.

[0063] MIC (Maximum Information Coefficient) reflects the degree of association between genes and cell types. For example, original genes with MIC below 0 can be removed, and the remaining original genes can be used as first-line candidate genes to reduce redundancy and non-informative features at the source. Specific algorithms for calculating MIC can be found in existing related technologies and will not be elaborated here.

[0064] Sub-step 2: For each first candidate gene, the gene expression vector corresponding to the first candidate gene is weighted based on the corresponding maximum information coefficient to obtain the weighted gene expression vector.

[0065] In this embodiment, the corresponding MIC can be obtained for each first candidate gene. Therefore, for example, the gene expression vector corresponding to each first candidate gene can be multiplied by its corresponding MIC to obtain the corresponding weighted gene expression vector.

[0066] Sub-step 3: Obtain the candidate gene expression matrix based on the weighted gene expression vectors corresponding to each first candidate gene.

[0067] In sub-step three, the candidate gene expression matrix can be obtained by combining the weighted gene expression vectors.

[0068] Sub-step four: Perform statistical analysis on the element values ​​in the candidate gene expression matrix, and determine the marker gene from the first candidate gene based on the statistical analysis results; the marker gene set is composed of marker genes.

[0069] In the above sub-step four, the F-score value (F-statistic) of each of the first candidate genes can be calculated based on the candidate gene expression matrix; and the marker gene can be determined from the first candidate genes based on the F-score value.

[0070] The F-score measures the ratio of between-class variance to within-class variance, assessing the distinguishing ability of a feature across different classes. In this embodiment, the F-score value for each first candidate gene is calculated using the following formula:

[0071]

[0072] in, This represents the F-score of the g-th first candidate gene. This indicates the number of single cells in plant type M. This represents the expression level of the g-th first candidate gene in a single cell of plant class M. This represents the average expression level of the g-th first candidate gene in all plant single cells, k is the number of plant single cell categories, and N is the total number of plant single cells. This represents the gene expression level of the i-th cell.

[0073] For example, the first candidate genes can be sorted in descending order of their F-score values; the top M second candidate genes with the highest F-score values ​​can be selected from the sorted first candidate genes; and the marker gene can be determined from the second candidate genes based on cross-validation. The specific value of M in this example can be flexibly set by the developers according to the actual situation.

[0074] In one example, K-fold cross-validation can be implemented. For each fold, the F-score is calculated for each first candidate gene. The top M candidate genes with the highest F-scores are selected as second candidate genes. Then, on the corresponding validation set, a classifier (such as logistic regression) is constructed using these second candidate genes, and its performance is evaluated, for example, the classification accuracy. This is repeated d times to obtain d accuracies. Simultaneously, the number of times each original gene is selected as a second candidate gene in K-fold cross-validation is recorded. The final marker gene is determined based on the frequency (stability) of the second candidate gene selection and the corresponding average classification accuracy.

[0075] In this example, the F-score statistical score is further calculated on the gene set determined based on MIC to refine the discriminative ranking and form candidate gene sequences from high to low, so as to enhance the effectiveness and stability of subsequent subset search.

[0076] In step S14 above, the cell subpopulation category to which each plant single cell belongs and the corresponding low-dimensional gene expression vector can be used as a training set and input into an initial model with a stacked ensemble structure for training to obtain a cell subpopulation classification model; the initial model includes multiple base models and a meta-model for fusing the initial prediction results of each base model to output the final prediction result.

[0077] The training process in this application embodiment may include:

[0078] In each round of cross-validation, each of the base models is trained on the first dataset of the training set, and the trained base models are used to predict the second dataset that was not trained, to obtain the prediction probability matrix of each base model for the second dataset. For each round, the prediction probability matrices are stacked column-wise to form a high-dimensional feature. The high-dimensional feature and the cell subpopulation category to which each of the plant single cells belongs are used as the input of the next layer meta-model. The meta-model learns to combine the prediction probability matrices of each base model to achieve dynamic weight allocation of the initial prediction results of each base model.

[0079] In one embodiment, the above method is applied to an electronic device that stores a cell subpopulation classification model corresponding to each organ of each plant species;

[0080] Before obtaining the low-dimensional cellular expression vector of the plant cells to be classified based on the marker gene set, the method includes:

[0081] The corresponding cell subpopulation classification model is invoked based on the received prediction instruction, which includes the species and organ to which the plant cell to be classified belongs.

[0082] To verify the effectiveness of the above methods, two sets of plant single-cell transcriptome data were obtained for experiments, from tobacco and rice. The two sets of data covered nine organs / developmental stages, totaling approximately 120,000 cells. Cross-tissue / cross-time evaluations were conducted on both species. Tobacco included tissues such as the corolla and included multi-timepoint tasks (including timepoint tri-classification and subgroup classification within each timepoint). Rice data covered eight organs / developmental stages and included large-scale cellular data for cross-tissue classification and gene selection generalization assessment.

[0083] Routine quality control was performed on cells and genes, retaining those with adequate expression levels and removing low-quality entries to obtain a gene expression matrix for analysis. The data was then divided into training and validation sets in an 8:2 ratio, with the training set used for feature selection and modeling, and the validation set used only for the final generalization evaluation.

[0084] Feature selection engineering (redundancy removal and optimal subset determination):

[0085] Maximum Information Coefficient (MIC) and Threshold Removal: Calculate the maximum information coefficient (MIC) between each original gene and the classification label, and remove genes with an MIC not higher than 0 to reduce redundant and non-informative features at the source.

[0086] Statistical scoring supplement: F-score values ​​are further calculated on the gene set obtained based on MIC screening to refine the discriminative ranking and form candidate gene sequences from high to low, so as to enhance the effectiveness and stability of subsequent subset search.

[0087] Incremental feature subset selection: Based on the above sorting, multiple candidate subsets are constructed in a forward incremental manner, and cross-validation is performed on the training set using XGBoost as the evaluator. The size of the target subset with the best overall performance is selected as the optimal set of marker genes (i.e., the marker genes mentioned above). For tobacco time point and subgroup tasks, about 60 genes are usually the optimal or near-optimal size. For rice cross-organ tasks, the optimal number of genes varies from about 60 to 610 depending on the organ (e.g., about 60 for crown root, about 110 for culm, about 160 for leaf, about 210 for flag leaf, about 610 for apical meristem, about 60 for tiller bud, about 110 for panicle, and about 210 for grain).

[0088] Stacked ensemble learning (cell subpopulation classification):

[0089] Base learner configuration: A stacked ensemble structure is adopted, with base learners selected from multiple models such as XGBoost, Support Vector Machine (SVM), Random Forest (RF), and LightGBM to ensure coverage of different types of decision boundaries and feature interactions.

[0090] Training and Fusion: Each base learner is trained on the optimal set of marker genes obtained above. After collecting their prediction outputs, they are stacked and fused to form the final ensemble classifier, which is used for cell subpopulation identification and robust prediction in cross-tissue / cross-species scenarios.

[0091] Evaluation and Comparison: Cross-validation was used during the training phase to stabilize the model performance. Finally, metrics such as accuracy, precision, and recall were reported on the independent test set. The comprehensive evaluation showed that, compared with a single learner, the above method generally improved the above metrics by about 1-3 percentage points, and achieved the same or better performance as whole-genome modeling with a smaller gene set.

[0092] Interpretability and biological consistency verification:

[0093] SHAP Contribution: The SHAP value is calculated for the trained ensemble classifier (base model + meta-model) to obtain the gene-level feature contribution ranking and candidate marker gene list, providing a basis for subsequent biological validation and functional annotation.

[0094] UMAP visualization: UMAP dimensionality reduction and clustering were performed using the optimal set of marker genes. The results showed that this subset can maintain or improve the separation of cell clusters and has a clearer structural presentation compared to whole-genome modeling.

[0095] PAGA Topological Consistency: A PAGA graph was constructed to compare the consistency of the optimal gene subset with the whole gene in terms of connectivity and developmental / differentiation trajectories, verifying that the subset can effectively preserve the biological landscape of the original data and support the rationality of the selected set of marker genes.

[0096] Preferably, SVM, LightGBM, and XGBoost were selected as the base models. The ROC curves fully demonstrate their superior performance. The ROC curves for SVM, LightGBM, and XGBoost as the base models in this experiment are shown in [reference needed]. Figure 5 As shown, this fully demonstrates its superior performance.

[0097] The online service module provides two types of services: "Search" and "Prediction". The search module allows users to browse marker gene resources and their contribution information by species, tissue and cell type. The prediction module allows users to upload expression matrices and call trained models to generate cell subpopulation classification results and candidate marker gene lists, while returning visualization reports based on SHAP, UMAP and PAGA.

[0098] It should be understood that although the steps in the flowchart above are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowchart above may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0099] In one embodiment, please refer to Figure 6 As shown, an electronic device is provided, including a processor 601 and a memory 602. The memory 602 stores a computer program, and the processor 601 executes the computer program to implement the steps of the method described above, which will not be repeated here.

[0100] Processor 601 can be an integrated circuit chip with signal processing capabilities. The processor 601 can be a general-purpose processor, including a CPU (Central Processing Unit), an NP (Network Processor), etc.; it can also be a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor.

[0101] The memory 602 may include, but is not limited to, RAM (Random Access Memory), ROM (Read Only Memory), PROM (Programmable Read Only Memory), EPROM (Erasable Programmable Read-Only Memory), and EEPROM (Electrically Erasable Programmable Read Only Memory).

[0102] Those skilled in the art will understand that Figure 6 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the application of the present application. Specific electronic devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0103] Based on the same inventive concept, embodiments of this application also provide a computer-readable storage medium, such as a floppy disk, optical disk, hard disk, flash memory, USB flash drive, SD (Secure Digital) card, MMC (Multi-Media Card), etc. The computer-readable storage medium stores a computer program, which, when executed by at least one processor, implements the steps of the methods in the above embodiments, which will not be repeated here.

[0104] Based on the same inventive concept, embodiments of this application also provide a computer program product, including a computer program that, when executed by a processor, implements any of the methods described above.

[0105] The program code for executing the computer program product of this application can be written in any combination of one or more programming languages. The program code can be executed entirely on the user device, partially on the user device, as a standalone software package, partially on the user device and partially on a remote device, or entirely on a remote device.

[0106] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, optical storage, etc.) containing computer-usable program code.

[0107] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer-readable storage media according to this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0108] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0109] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of user-operated steps to be executed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1The steps of the function specified in one or more boxes.

[0110] It should be noted that the illustrations provided in this embodiment are only schematic representations of the basic concept of this application. Therefore, the drawings only show components relevant to this application and are not drawn according to the actual number, shape, and size of components in implementation. In actual implementation, the shape, quantity, and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex. The structures, proportions, sizes, etc., shown in the accompanying drawings are only used to complement the content disclosed in the specification for those skilled in the art to understand and read, and are not intended to limit the implementation conditions of this application. Therefore, they have no substantial technical significance. Any modification to the structure, change in the proportional relationship, or adjustment of the size, without affecting the effect and purpose that this application can produce, should still fall within the scope of the technical content disclosed in this application. At the same time, the terms such as "upper," "lower," "left," "right," "middle," and "one" used in this specification are only for clarity of description and are not intended to limit the scope of implementation of this application. Changes or adjustments in their relative relationships, without substantially changing the technical content, should also be considered within the scope of implementation of this application.

[0111] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the document does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0112] As illustrated herein, unless the context clearly indicates otherwise, the words “a,” “an,” “an,” and / or “the” do not specifically refer to the singular and may also include the plural. Generally speaking, the terms “comprising” and “including” only indicate the inclusion of explicitly identified steps and elements, which do not constitute an exclusive list, and the method or apparatus may also include other steps or elements.

[0113] The definitions used herein, such as the terms “having,” “may have,” “comprising,” or “may include,” indicate the presence of the corresponding function, operation, element, etc., and do not limit the presence of one or more other functions, operations, elements, etc. Furthermore, it should be understood that the terms “comprising” or “having” as used herein indicate the presence of the features, figures, steps, operations, elements, components, or combinations thereof described in the specification, without excluding the presence or addition of one or more other features, figures, steps, operations, elements, components, or combinations thereof.

[0114] The prefixes such as "first" and "second" used in this application embodiment are merely for distinguishing different descriptive objects and do not limit the position, order, priority parameters, quantity, or content of the described objects. The use of ordinal numbers and other prefixes used to distinguish descriptive objects in this application embodiment does not constitute a limitation on the described objects. For statements regarding the described objects, please refer to the claims or the context of the embodiments. The use of such prefixes should not constitute unnecessary limitations. Furthermore, in the description of this embodiment, unless otherwise stated, "multiple" means two or more.

[0115] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0116] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for classifying cell subpopulations based on plant single-cell transcriptome marker genes, characterized in that, The method comprises the following steps: obtaining plant single-cell transcriptome sequencing data; the plant single-cell transcriptome sequencing data comprises a gene expression matrix corresponding to a plurality of plant single cells and a cell subpopulation category to which each plant single cell belongs, and each element value in the gene expression matrix represents the expression amount of a corresponding original gene in a corresponding plant single cell; eliminating original genes with low-quality expression based on the gene expression matrix to obtain a marker gene set; determining a low-dimensional cell expression vector corresponding to each plant single cell based on the marker gene set; training a cell subpopulation classification model based on the low-dimensional cell expression vector corresponding to each plant single cell and the cell subpopulation category to which each plant single cell belongs; obtaining a low-dimensional cell expression vector of a plant cell to be classified based on the marker gene set; determining the cell subpopulation to which the plant cell to be classified belongs according to the low-dimensional cell expression vector of the plant cell to be classified and the cell subpopulation classification model.

2. The method for classifying cell subpopulations based on plant single-cell transcriptome marker genes according to claim 1, characterized in that, eliminating original genes with low-quality expression based on the gene expression matrix to obtain a marker gene set, comprising: obtaining a gene expression vector corresponding to each original gene based on the gene expression matrix; the gene expression vector is a sequence composed of the expression amounts of the original gene in each plant single cell; for each original gene, calculating the maximum information coefficient between the original gene and the category label according to the gene expression vector; eliminating original genes with low-quality expression according to the maximum information coefficient to obtain a marker gene set.

3. The method for classifying cell subpopulations based on plant single-cell transcriptome marker genes according to claim 2, characterized in that, eliminating original genes with low-quality expression according to the maximum information coefficient to obtain a marker gene set, comprising: eliminating original genes with a maximum information coefficient lower than a preset threshold to obtain first candidate genes; for each first candidate gene, weighting the gene expression vector corresponding to the first candidate gene based on the corresponding maximum information coefficient to obtain a weighted gene expression vector; obtaining a candidate gene expression matrix according to the weighted gene expression vector corresponding to each first candidate gene; statistically analyzing the element values in the candidate gene expression matrix, and determining a marker gene from the first candidate genes according to the statistical analysis result; the marker gene set is composed of the marker gene.

4. The method for classifying cell subpopulations based on plant single-cell transcriptome marker genes according to claim 3, characterized in that, statistically analyzing the element values in the candidate gene expression matrix, and determining a marker gene from the first candidate genes according to the statistical analysis result, comprising: calculating the F-score value of each first candidate gene according to the candidate gene expression matrix; determining a marker gene from the first candidate genes according to the F-score value.

5. The method for classifying cell subpopulations based on plant single-cell transcriptome marker genes according to claim 4, wherein, determining a marker gene from the first candidate genes according to the F-score value, comprising: sorting each first candidate gene in descending order of F-score value; selecting the top M second candidate genes from the sorted first candidate genes in terms of F-score value, and determining a marker gene from the second candidate genes based on cross-validation.

6. The method for classifying cell subpopulations based on plant single-cell transcriptome marker genes according to claim 1, wherein, training a cell subpopulation classification model based on the cell subpopulation category to which each plant single cell belongs and the corresponding low-dimensional gene expression vector, comprising: The cell subpopulation category to which each of the plant single cells belongs and the corresponding low-dimensional gene expression vector are input into an initial model adopting a stacked integration structure for training to obtain a cell subpopulation classification model; the initial model includes a plurality of base models and a meta-model for fusing initial prediction results of each of the base models to output a final prediction result.

7. The method for classifying cell subpopulations based on plant single-cell transcriptome marker genes according to claim 6, wherein, The training process includes: In each round of cross-validation, each of the base models is trained on a first data set of the training set, and each of the trained base models is used to predict a second data set that does not participate in training to obtain a prediction probability matrix of each of the base models on the second data set; For each round, each of the prediction probability matrices is stacked by column to form a high-dimensional feature; The high-dimensional feature and the cell subpopulation category to which each of the plant single cells belongs are jointly input into a next layer meta-model, and the meta-model learns to combine the prediction probability matrices of each of the base models to achieve dynamic weight distribution of the initial prediction results of each base model.

8. The method for classifying cell subpopulations based on plant single-cell transcriptome marker genes according to claim 1, wherein, The method is applied to an electronic device, and the electronic device stores cell subpopulation classification models corresponding to each organ of each plant species; Before obtaining a low-dimensional cell expression vector of a plant cell to be classified based on the marker gene set, the method includes: According to the received prediction instruction, the corresponding cell subpopulation classification model is called, and the prediction instruction includes the species and organ to which the plant cell to be classified belongs.

Images