Megakaryocyte identification method and system based on single cell transcriptome

Abstract

The application provides a megakaryocyte recognition method and system based on single-cell transcriptome, comprising: obtaining a single-cell transcriptome expression matrix to be tested and performing quality control and data preprocessing, extracting expression information of megakaryocyte standard genes in the preprocessed data to generate a feature vector of the cell to be tested; according to a pre-trained machine learning classification model, classifying each cell to be tested as a megakaryocyte, and outputting a prediction result of the cell to be tested; based on a pre-set double judgment strategy, comparing the prediction result output by the model with a pre-set standard gene expression proportion threshold to obtain a predicted megakaryocyte recognition result. The application integrates single-cell transcriptomes of megakaryocytes in multiple organs and multiple development stages, constructs a standard megakaryocyte gene set and trains a machine learning classification model, thereby improving the recognition accuracy and transferability of megakaryocytes in traditional known organs and organs that have not been fully recognized in the past.

Description

Technical Field

[0001] This application belongs to the field of machine learning technology, and in particular relates to a method and system for megakaryocyte recognition based on single-cell transcriptome. Background Technology

[0002] Megakaryocytes (MKs) are large polyploid cells in the bone marrow that produce platelets. They are commonly identified using a few classic biomarkers such as Pf4, Itga2b, Gp9, and Itgb3. However, single or a few classic biomarkers have insufficient specificity and sensitivity in complex single-cell data. For example, CD41 / CD61 can also be found in hematopoietic progenitor / precursor cells, mast cells, or other cell types; megakaryocytes from different organs and at different developmental stages also exhibit significant transcriptomic differences, making manual annotation based on a fixed set of few biomarkers prone to missed or false positives.

[0003] With the rapid accumulation of single-cell atlases of multiple tissues in mice and humans, covering various organs from the embryonic, neonatal, and adult stages, megakaryocytes are often overlooked or categorized as platelets, myeloid cells, endothelial-associated cells, or unannotated populations in general annotation workflows due to their scarcity, large size, fragility, wide range of tissue origins, and diverse transcriptional states. Traditional histological, immunostaining, or manual screening methods struggle to systematically discover the distribution patterns of megakaryocytes across organs and developmental stages within millions of single-cell datasets.

[0004] Existing general-purpose single-cell annotation tools (such as automatic annotation methods based on reference datasets or gene sets) are mainly geared towards common cell types and lack specific optimization for the identification of rare and functionally heterogeneous megakaryocytes. While relying solely on the classic four-gene combination can reduce false positives, it is easy to sacrifice the true positive rate; on the other hand, simply relaxing the marker threshold may introduce a large number of false positives from non-megakaryocytes.

[0005] Therefore, there is an urgent need for a single-cell transcriptome megakaryocyte recognition system that combines a standard gene set of multi-organ megakaryocytes with a supervised learning model. Summary of the Invention

[0006] In view of this, this application aims to propose a megakaryocyte recognition method and system based on single-cell transcriptomics to solve at least one of the above-mentioned problems.

[0007] To achieve the above objectives, the technical solution of this application is implemented as follows: In a first aspect, this application provides a megakaryocyte recognition method based on single-cell transcriptomics, including: The transcriptome expression matrix of the single cell to be tested was obtained and quality control and data preprocessing were performed. The expression information of standard genes of megakaryocytes in the preprocessed data was extracted to generate the feature vector of the cell to be tested. Based on the pre-trained machine learning classification model, each cell to be tested is classified as a megakaryocyte, and the prediction result of the cell to be tested is output; wherein, the machine learning classification model is obtained by dividing the expression matrix of single cells labeled as megakaryocytes and non-megakaryocytes into training set and test set according to a preset ratio for training and testing. Based on a pre-defined dual-judgment strategy, the predicted results output by the model are compared with the preset standard gene expression ratio threshold to obtain the predicted megakaryocyte recognition results.

[0008] Secondly, based on the same inventive concept, this application also provides a megakaryocyte recognition system based on single-cell transcriptome, comprising: The data acquisition and processing module is configured to acquire the transcriptome expression matrix of the single cell to be tested and perform quality control and data preprocessing, extract the expression information of megakaryocyte standard genes in the preprocessed data, and generate the feature vector of the cell to be tested. The model prediction module is configured to classify each cell to be tested as a megakaryocyte by calling a pre-trained machine learning classification model and output the prediction result of the cell to be tested; wherein, the machine learning classification model is obtained by dividing the expression matrix of single cells labeled as megakaryocytes and non-megakaryocytes into training set and test set according to a preset ratio for training and testing. The judgment and recognition module is configured to compare the predicted results output by the model with the preset standard gene expression ratio threshold based on a pre-set dual judgment strategy in order to obtain the predicted megakaryocyte recognition results.

[0009] Thirdly, based on the same inventive concept, this application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method described in the first aspect.

[0010] Fourthly, based on the same inventive concept, this application also provides a non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores computer instructions for causing the computer to perform the method as described in the first aspect.

[0011] Compared with existing technologies, the megakaryocyte recognition method and system based on single-cell transcriptome described in this application have the following advantages: The megakaryocyte identification method based on single-cell transcriptomics described in this application uses real megakaryocyte transcriptomics from multiple organs and stages as the training basis, constructs a feature set by combining classic and newly discovered megakaryocyte marker genes, and uses supervised learning models such as linear support vector machines for classification. At the same time, it combines a dual judgment strategy of "the model judges it as a megakaryocyte and expresses a standard gene at a rate not lower than a preset proportion" to reduce false positives and improve the ability to identify rare cells, thereby improving the accuracy and transferability of megakaryocyte identification in traditionally known organs and organs that have not been fully understood in the past. Attached Figure Description

[0012] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a flowchart of a megakaryocyte recognition method based on single-cell transcriptomics as described in an embodiment of this application; Figure 2 This is a schematic diagram of the standard gene set for megakaryocytes described in the embodiments of this application; Figure 3 This is a diagram showing the expression of the novel megakaryocyte markers, including TNIK, described in the embodiments of this application in megakaryocytes of multiple organs and platelets of peripheral blood. Figure 4 This is a graph showing the recognition performance of the classic four-marker scheme described in the embodiments of this application in representative labeled data; Figure 5 The ROC curves of the 26-gene MKIDS model described in this application embodiment are shown in the internal and external test sets. Figure 6 This is a comparison chart of the F1-scores of the 10-gene combination containing biomarkers described in the embodiments of this application and the classic biomarker scheme; Figure 7 This is a comparison chart of the F1-scores of the 26-gene combination containing biomarkers described in the embodiments of this application and the classic four-biomarker scheme. Figure 8 This is a verification diagram of the Pf4-Cre-mediated Tnik conditional knockout model described in the embodiments of this application; Figure 9 This is a diagram illustrating how Tnik deficiency affects pre-platelet formation, sinus localization, and terminal platelet output in megakaryocytes, as described in the embodiments of this application. Figure 10 This is a diagram showing the application results of the classification model described in the embodiments of this application in single-cell data from multiple organs and multiple developmental stages; Figure 11 This is a schematic diagram of a megakaryocyte recognition system based on single-cell transcriptomics as described in an embodiment of this application; Figure 12 This is a schematic diagram of the hardware structure of the electronic device described in an embodiment of this application. Detailed Implementation

[0013] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with specific embodiments and the accompanying drawings.

[0014] It should be noted that, unless otherwise defined, the technical or scientific terms used in the embodiments of this application should have the ordinary meaning understood by one of ordinary skill in the art to which this application pertains. The terms "first," "second," and similar terms used in the embodiments of this application do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed after the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are only used to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0015] The embodiments of this application are described in detail below with reference to the accompanying drawings.

[0016] Please see Figure 1 As shown, this embodiment provides a megakaryocyte recognition method based on single-cell transcriptomics, which specifically includes the following steps: Step S101: Obtain the transcriptome expression matrix of the single cell to be tested and perform quality control and data preprocessing. Extract the expression information of standard genes of megakaryocytes from the preprocessed data to generate the feature vector of the cell to be tested. Step S102: According to the pre-trained machine learning classification model, classify each cell to be tested as a megakaryocyte and output the prediction result of the cell to be tested; wherein, the machine learning classification model is obtained by dividing the single cell expression matrix labeled as megakaryocyte and non-megakaryocyte into training set and test set according to a preset ratio for training and testing. Step S103: Based on the pre-set dual judgment strategy, compare the prediction results output by the model with the preset standard gene expression ratio threshold to obtain the predicted megakaryocyte recognition results.

[0017] The identification method described in this embodiment takes a single-cell gene expression matrix as input. After quality control, normalization, feature gene extraction, machine learning classification, and standard gene expression rule determination, it outputs the predicted megakaryocyte (pMKs) identification results. Addressing the specific problems of megakaryocytes being rare, easily missed, and exhibiting strong cross-organ heterogeneity in single-cell transcriptome data, the method establishes a standard gene set of 10 megakaryocytes based on real megakaryocytes from multiple organs and developmental stages. Its core includes classic megakaryocyte markers as well as novel markers, including Tnik, Dlg2, Pard3b, Ubash3b, Ank3, and Kalrn. It combines a non-megakaryocyte negative training set, a linear support vector machine classifier, and a standard gene expression ratio threshold for dual determination, thereby improving the sensitivity, specificity, and transferability of predicted megakaryocyte identification.

[0018] This method integrates multi-tissue standard gene sets with machine learning classification models to achieve accurate identification of megakaryocytes (MKs) in single-cell data. Compared with the classic four-gene combination of Pf4, Itga2b, Gp9, and Itgb3, it can improve sensitivity, specificity, and F1-score in multiple internal and external datasets, reduce false negatives caused by traditional biomarker strategies, and effectively improve identification accuracy.

[0019] In some implementations, single-cell transcriptome data are collected from known megakaryocyte-related organs and developmental stages; wherein known megakaryocytes are obtained through flow cytometry sorting, published single-cell type annotation, or combination of markers. Single-cell transcriptome data underwent quality control and data preprocessing. Genes that were stably and highly expressed in megakaryocytes across multiple organs and stages and were distinguishable from non-megakaryocytes were screened to construct a standard gene set for megakaryocytes.

[0020] Specifically, in this embodiment, single-cell transcriptome data from known megakaryocyte-related organs and developmental stages were collected, including but not limited to tissues such as yolk sac, peripheral blood, fetal liver / liver, lung, spleen, and bone marrow, covering the embryonic to adulthood period. Known megakaryocytes can be obtained through FACS sorting, published single-cell annotation, or combined biomarkers. Megakaryocytes confirmed by flow cytometry sorting or literature annotation were used as a positive training set. Simultaneously, annotated non-megakaryocytes from published single-cell data on the same platform were collected as negative control data, including GSE81682, GSE67120, GSE116530, GSE167588, and GSE94877. Based on the raw data or publicly available annotation results, various cell types annotated as non-megakaryocytes were selected as negative samples. Non-megakaryocyte types include erythroid cells, B cells, T cells, NK cells, monocytes, macrophages, granulocytes, endothelial cells, hematopoietic stem cells, hematopoietic progenitor cells, megakaryocyte-erythroid progenitor cells, granulocyte-monocyte progenitor cells, and lymphocyte progenitor cells. These non-megakaryocytes are used in the standard gene screening phase to exclude non-specifically expressed genes.

[0021] The data undergoes unified preprocessing (including gene name unification, low-quality cell filtering, low-expression gene filtering, logarithmic transformation, batch integration, or standardization). Quality control indicators may include the number of genes detected in each cell, UMI number, mitochondrial gene ratio, ribosomal gene ratio, etc. Empty droplets or low-quality cells with too low gene or UMI number are removed, as are suspected damaged cells beyond median ±2SD of mitochondrial gene expression.

[0022] Genes with stable and high expression levels, and distinguishable from non-megakaryocytes, were screened in megakaryocytes from multiple organs and stages. Marker genes were ranked for each tissue-derived megakaryocyte type, and the top 500 genes with the highest average expression levels in each tissue-derived megakaryocyte type were selected. The median rank of each gene across different tissue-derived megakaryocyte types was calculated. Simultaneously, marker genes were ranked for each non-megakaryocyte type, and the top 500 genes with the highest average expression levels in each non-megakaryocyte type were selected. The median rank of each gene across different non-megakaryocyte types was calculated. 175 overlapping genes were removed, and the top 80% of genes with a fold increase greater than 2 in expression levels within megakaryocytes compared to other cell types were further screened. The results are as follows: Figure 2 The standard gene set for megakaryocytes shown includes classic megakaryocyte markers Pf4, Itga2b, Gp9, and Itgb3, as well as newly identified or previously underutilized markers for megakaryocyte recognition, such as Tnik, Dlg2, Pard3b, Ubash3b, Ank3, and Kalrn.

[0023] For validation of the expression of the novel biomarkers, single-cell transcriptome analysis showed that the above six novel biomarkers were stably expressed in megakaryocytes across organs but expressed at low levels in non-megakaryocytes; such as Figure 3 As shown, immunofluorescence further verified the expression of TNIK, DLG2, UBASH3B, ANK3, and KALRN in CD41+CD42d+ megakaryocytes derived from the yolk sac, peripheral blood, liver, lung, spleen, and bone marrow, with lower expression in non-megakaryocytes of the bone marrow; the above proteins can also be detected in adult peripheral blood platelets.

[0024] This embodiment is applicable across organs and developmental stages, and can be used for known megakaryocyte organs such as yolk sac, peripheral blood, liver, lung, spleen, and bone marrow, as well as for megakaryocyte identification in candidate organs such as brain, heart, placenta, pancreas, gonads, stomach, and aorta.

[0025] In some implementations, methods for constructing machine learning classification models include: The single-cell expression matrices labeled as megakaryocytes and non-megakaryocytes are divided into training and testing sets according to a preset ratio. Each cell is composed of a feature vector consisting of the expression level, expression ratio, model score, or a combination thereof of the standard genes of megakaryocytes. A binary classification model is constructed using linear support vector machines, and the model parameters are optimized through cross-validation.

[0026] Specifically, in this embodiment, the single-cell expression matrices labeled as megakaryocytes and non-megakaryocytes are divided into training and testing sets according to a preset ratio. Approximately 200 megakaryocytes are selected from each of the six known megakaryocyte-related organs, totaling approximately 1200 megakaryocytes as positive samples. Simultaneously, non-megakaryocytes are selected as negative samples from the non-megakaryocyte negative control data, with approximately 100 cells selected for each non-megakaryocyte type. For rare cell types with fewer than 100 samples, all usable cells can be selected. This results in a negative set of approximately 1248 non-megakaryocytes, used as negative training samples in the MKIDS binary classification model training. The training set:test set ratio is 9:1.

[0027] For training the classifier, it is preferable to use a linear support vector machine (linear SVM) to build a binary classification model; the SVM type is C-classification, the kernel function is linear, and parameters such as cost and gamma are selected through cross-validation.

[0028] Meanwhile, k-fold cross-validation was adopted, with 10-fold cross-validation being preferred. To reduce the fluctuation caused by random partitioning, the cross-validation was repeated multiple times, up to 100 times. One or more of the following were used as model selection indicators: specificity, sensitivity, F1-score, AUC, accuracy, positive predictive value, and negative predictive value.

[0029] It should be noted that the classifier can be replaced by a linear SVM, or by a non-linear SVM, logistic regression, random forest, gradient boosting tree, XGBoost, LightGBM, neural network, Naive Bayes, k-nearest neighbors, ensemble learning, or deep learning model.

[0030] For the cells to be tested, the machine learning classifier first outputs the megakaryocyte / non-megakaryocyte category (or predicted probability). When the cell is identified as a megakaryocyte by the model, and the number or proportion of megakaryocyte standard genes expressed by the cell reaches a preset threshold, it is identified as a predicted megakaryocyte, thereby effectively eliminating false positive cells caused by data noise or model overfitting.

[0031] Specifically, for the cells to be tested, the standardized expression matrices of 10 megakaryocyte standard genes are first extracted and input into the trained MKIDS classifier to obtain the MK / non-MK classification results and corresponding model judgment scores for the cells. Then, the number or proportion of megakaryocyte standard genes expressed above a preset expression threshold is calculated. A cell is defined as a predicted megakaryocyte (pMKs) only if it simultaneously meets the following conditions: 1) the classifier classifies it as MK; 2) the proportion of megakaryocyte standard genes expressed by the cell to the total number of standard genes is not less than 50%. If only one condition is met, it is not classified as a pMK.

[0032] It should be noted that this embodiment selects the expression of at least 50% of the megakaryocyte standard gene as the preferred threshold. However, the threshold can also be set to a range of 30%-70% depending on the data depth, species, platform, or tissue source. Alternatively, dynamic thresholds such as probability thresholds, module score thresholds, quantile thresholds, and ROC optimal cutoff points can be used. This embodiment is only used as an example for illustration and is not specifically limited here.

[0033] This embodiment employs a dual judgment strategy of "machine learning classification + standard gene expression ratio threshold" to improve the detection rate of rare megakaryocytes while reducing false positives of non-megakaryocytes.

[0034] Furthermore, this embodiment employs a random forest feature importance assessment to evaluate the contribution of each gene to the recognition performance, thereby outputting key discriminant genes and their weights. Specifically, for importance assessment, this embodiment uses the pre-trained SVM model's prediction results as a basis, refits the data using a random forest algorithm, and calculates the feature importance score of each gene by evaluating the contribution of each standard gene to purity improvement (such as the Gini coefficient or information gain) during tree model node splitting.

[0035] For the key discriminant genes, the standard gene set is ranked according to importance scores, and the top-ranked core discriminant genes and their corresponding weights are output. This not only verifies whether the model has captured known biological markers (such as Pf4, Itga2b), but may also discover new key driver genes.

[0036] Based on the above method, this embodiment provides the following example for verification: Example 1: Comparison of recognition performance between a 10-gene combination containing new biomarkers and a classic four-biomarker scheme.

[0037] (1) Comparison scheme: The classic biomarker scheme identifies cells simultaneously expressing Pf4, Itga2b, Gp9, and Itgb3 as megakaryocytes; the new biomarker scheme extracts the expression values ​​of 10 standard genes, including Tnik, Dlg2, Pard3b, Ubash3b, Ank3, and Kalrn, inputs them into a linear SVM classifier, and combines the rule that the expression ratio of standard genes is not less than 50% (for the 10 standard gene expression values, the standard gene expression ratio is: when the normalized and logarithmically transformed expression value of a gene is greater than 0, the gene is considered to be expressed, and the expression ratio of not less than 50% means that at least 5 standard gene expression values ​​are >0) to obtain the final predicted megakaryocytes. The two schemes are compared on the same internal test set and four independent external test sets.

[0038] (2) Identification results of the new marker combination: such as Figure 6 As shown, the F1-scores of the 10-gene combination containing the novel biomarker were 0.9960, 0.9344, 0.6728, and 0.6985 in the internal and external test sets, respectively. In almost all test sets, the 10-gene combination containing the novel biomarker, along with the dual-determination strategy (where the machine learning classification result and the standard gene expression ratio >50% are simultaneously satisfied), outperformed the classic four-biomarker approach in terms of sensitivity, specificity, and F1-score. This indicates that the combination can maintain a low false positive rate while reducing false negatives caused by the classic biomarker approach.

[0039] Example 2: Comparison of recognition performance between the 26-gene combination containing the new biomarker and the classic four-biomarker scheme.

[0040] (1) Comparison scheme: The classic biomarker scheme identifies cells that simultaneously express Pf4, Itga2b, Gp9, and Itgb3 as megakaryocytes; the new biomarker scheme extracts the expression values ​​of 26 standard genes including Tnik, Dlg2, Pard3b, Ubash3b, Ank3, and Kalrn (the feature vector of each cell is composed of the normalized expression values ​​of the 26 standard megakaryocyte genes (the normalized expression value refers to the value obtained after library depth normalization, logarithmic transformation, and feature scale unification of the original gene expression count of a single cell, which is used as the input feature of the machine learning classifier). After normalization and logarithmic transformation, Pf4 (platelet factor 4), Ppbp (pro-platelet basic protein), Myl9 (myosin, lightpolypeptide 9, regulatory), Thbs1 (thrombospondin 1), Gp6 (glycoprotein 6platelet), Alox12 (arachidonate 12-lipoxygenase), and Itgb3 (integrin beta) are extracted. 3), Itga2b (integrin alpha 2b), Rab27b (RAB27B, member RAS oncogene family), Pde5a (phosphodiesterase 5A, cGMP-specific), Gp9 (glycoprotein 9 platelet), Treml1 (triggering receptor expressed on myeloid cells-like 1), Clec1b (C-type lectindomain family 1, member b), Clu (clusterin), Mrvi1 (murine retrovirus integration site 1 homolog), Parvb (parvin,beta), Gp5 (glycoprotein 5 platelet), Daam1 (dishevelled associated activator of morphogenesis 1), Rbpms2 (RNA bindingprotein with multiple splicing 2), Lims1 (LIM and senescent cell antigen-likedomains 1), Dlg2 (discs large MAGUK scaffold protein 2), Pard3b (par-3 family cell polarity regulator beta), Ubash3b (ubiquitin associated and SH3 domaincontaining, B), Ank3 (ankyrin 3, epithelial), Kalrn (kalirin, RhoGEF kinase) and Tnik (TRAF2 and NCK interacting The expression level of kinase in single cells was input into a linear SVM classifier, and combined with a rule that the expression ratio of standard genes should not be less than 50% (for 26 standard gene expression values, the standard gene expression ratio is: when the normalized and logarithmically transformed expression value of a gene is greater than 0, the gene is considered to be expressed; an expression ratio of not less than 50% means that at least 13 standard gene expression values ​​are >0) to obtain the final predicted megakaryocyte. The two schemes were compared on the same internal test set and four independent external test sets.

[0041] (2) Limitations of the classic four-marker scheme: such as Figure 4 As shown, in representative known labeled data, using Pf4, Itga2b, Gp9 and Itgb3 simultaneously can control the false positive rate to below 1%, but the true positive rate is only 80.35%, that is, about 19.65% of real megakaryocytes are not identified due to classical gene dropout, low expression or organ / developmental stage heterogeneity.

[0042] (3) Identification results of the new marker combination: such as Figure 5 As shown, the AUCs of MKIDS on the internal and external test sets were 0.9992, 0.9977, 0.8888, 0.9997, and 0.9265, respectively, indicating that the model can effectively distinguish between megakaryocytes and non-megakaryocytes and maintains high recognition accuracy and external generalization ability in single-cell transcriptome data from different sources. Figure 7As shown, the F1-scores of MKIDS on the internal and external test sets were 0.9858, 0.9989, 0.8597, and 0.7291, respectively. In almost all test sets, the 26-gene combination containing the novel biomarker and the dual-determination strategy outperformed the classic four-biomarker approach in terms of sensitivity, specificity, and F1-score, indicating that this combination can maintain a low false positive rate while reducing false negatives caused by the classic biomarker approach.

[0043] Example 3: Functional validation of the novel biomarker Tnik in preplatelet formation and platelet production in megakaryocytes.

[0044] Expression and conditional knockout: After validating TNIK expression in CD41+CD42d+ megakaryocytes and peripheral blood platelets in multiple organs, Tnik flox mice were constructed, such as... Figure 8 As shown, Figure 8 A represents megakaryocyte / platelet-specific Tnik-deficient mice (Pf4-Tnik) obtained by deleting exons 3-6 of Tnik using Pf4-Cre. - / - ), with Tnik from the same nest fl / fl Mice served as a control. Figure 8 B and 8C show Pf4-Tnik in immunofluorescence and platelet Western blot. - / - TNIK signaling was significantly reduced or nearly absent in mouse bone marrow megakaryocytes and circulating platelets.

[0045] like Figure 9 This diagram illustrates how Tnik deficiency affects pre-megakaryocyte platelet formation, sinus localization, and terminal platelet output. Figure 9 Image A shows three-dimensional imaging and surface reconstruction results of megakaryocytes in situ in the bone marrow. CD41 was used to label megakaryocytes, endomucin was used to label bone marrow vascular sinuses, and DAPI was used to label cell nuclei. Tnik fl / fl In the control group, megakaryocyte formation and extension of preplatelet-like structures were observed, while in Pf4-Tnik... - / - The knockout group showed a reduction in megakaryocytes that form pre-platelet structures. Quantitative results showed that Pf4-Tnik... - / - The proportion of megakaryocytes in the knockout group before platelet formation was significantly lower than that in the Tnik group. fl / fl Control group. Figure 9 B shows the preplatelet formation of sorted bone marrow megakaryocytes after in vitro culture. CD41 was used to label megakaryocytes, α-tubulin and F-actin were used to visualize the megakaryocyte cytoskeleton and preplatelet processes, and DAPI was used to label the cell nucleus. fl / fl In the control group, megakaryocytes could form longer, branched proplatelet processes, while Pf4-Tnik... - / -The knockout group showed reduced proplatelet extension and branching in megakaryocytes; quantitative results indicated that Pf4-Tnik... - / - The proportion of pre-platelet formation in the group was significantly reduced. Figure 9 C shows the spatial relationship between bone marrow megakaryocytes and endomucin-positive vascular sinuses. Megakaryocytes are divided into groups no larger than 1000 μm based on their surface area. 2 1000~2000μm 2 and greater than 2000μm 2 Three intervals were used, and the shortest distance between the megakaryocyte and the nearest vascular sinus was calculated. Results showed that in areas with a surface area not exceeding 1000 μm... 2 In megakaryocytes, there was no significant difference in the distance between the two groups and the vascular sinuses; in cells with a surface area of ​​1000–2000 μm 2 and greater than 2000μm 2 Pf4-Tnik - / - The shortest distance between the group and the vascular sinus was significantly increased, indicating that Tnik deficiency mainly affects the ability of larger megakaryocytes to approach or locate towards the vascular sinus. Figure 9 D shows the results of peripheral blood platelet parameter testing. (Compared to Tnik) fl / fl Compared with the control group, Pf4-Tnik - / - Mice in the knockout group showed significantly reduced peripheral blood platelet counts and plateletcrit (PCT). Combined with in vitro and in vivo results of proplatelet formation and vascular sinus localization, this indicates that Tnik deficiency impairs megakaryocyte proplatelet formation and terminal platelet output.

[0046] Based on the above method, this embodiment also provides specific application examples: Example 4: Application in single-cell data of multiple organs in mice.

[0047] The trained model is then applied to a large-scale single-cell mouse dataset, such as... Figure 10 As shown, the data covers 31 organs and stages including embryonic, neonatal, and adulthood. For each organ / stage, the system outputs predicted megakaryocyte numbers, predicted proportions, standard gene expression, transcriptomic similarity to known megakaryocytes, and enrichment results for megakaryocyte-related functions such as coagulation, hemostasis, platelet activation, and wound healing. This application has discovered candidate organs for megakaryocyte residency beyond traditionally known organs, such as the brain, heart, placenta, pancreas, gonads, stomach, prostate, and aorta.

[0048] Example 5: Human-derived classification model and its cross-species application.

[0049] In human single-cell data, a human classification model can be constructed based on the expression matrices of human megakaryocytes and non-megakaryocytes. Human megakaryocytes and non-megakaryocytes were selected as training data, with the training and test sets divided in a 9:1 ratio. A linear SVM model was employed, and parameters were optimized through cross-validation. The human classification model automatically identified megakaryocytes in human embryonic / fetal and adult multi-tissue single-cell atlases and output the organ distribution of human megakaryocytes.

[0050] It should be noted that the above description describes some embodiments of this application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in a different order than that shown in the above embodiments and still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0051] Based on the same inventive concept, and corresponding to any of the above embodiments, the embodiments of this application also provide a megakaryocyte recognition system based on single-cell transcriptomics.

[0052] like Figure 11 As shown, the megakaryocyte recognition system based on single-cell transcriptomics includes: The data acquisition and processing module 11 is configured to acquire the transcriptome expression matrix of the single cell to be tested and perform quality control and data preprocessing, extract the expression information of megakaryocyte standard genes in the preprocessed data, and generate the feature vector of the cell to be tested. The model prediction module 12 is configured to classify each cell to be tested as a megakaryocyte by calling a pre-trained machine learning classification model and output the prediction result of the cell to be tested; wherein, the machine learning classification model is obtained by dividing the single-cell expression matrix labeled as megakaryocytes and non-megakaryocytes into a training set and a test set according to a preset ratio for training and testing. The determination and recognition module 13 is configured to compare the predicted results output by the model with the preset standard gene expression ratio threshold based on a pre-set dual determination strategy, so as to obtain the predicted megakaryocyte recognition results.

[0053] For ease of description, the above apparatus is described in terms of its functions, divided into various modules. Of course, in implementing the embodiments of this application, the functions of each module can be implemented in one or more software and / or hardware.

[0054] The apparatus of the above embodiments is used to implement the corresponding method in any of the foregoing embodiments and has the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0055] Based on the same inventive concept, corresponding to the methods of any of the above embodiments, embodiments of this application also provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the methods described in any of the above embodiments.

[0056] Figure 12 This embodiment illustrates a more specific hardware structure of an electronic device, which may include a processor 1010, a memory 1020, an input / output interface 1030, a communication interface 1040, and a bus 1050. The processor 1010, memory 1020, input / output interface 1030, and communication interface 1040 are interconnected internally via the bus 1050.

[0057] The processor 1010 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification.

[0058] The memory 1020 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 1020 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented by software or firmware, the relevant program code is stored in the memory 1020 and is called and executed by the processor 1010.

[0059] The input / output interface 1030 is used to connect input / output modules to realize information input and output. The input / output modules can be configured as components in the device (not shown in the figure) or externally connected to the device to provide corresponding functions. Input devices may include keyboards, mice, touch screens, microphones, various sensors, etc., and output devices may include displays, speakers, vibrators, indicator lights, etc.

[0060] The communication interface 1040 is used to connect a communication module (not shown in the figure) to enable communication between this device and other devices. The communication module can communicate via wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0061] Bus 1050 includes a pathway for transmitting information between various components of the device, such as processor 1010, memory 1020, input / output interface 1030, and communication interface 1040.

[0062] It should be noted that although the above-described device only shows the processor 1010, memory 1020, input / output interface 1030, communication interface 1040, and bus 1050, in specific implementations, the device may also include other components necessary for normal operation. Furthermore, those skilled in the art will understand that the above-described device may only include the components necessary for implementing the embodiments of this specification, and not necessarily all the components shown in the figures.

[0063] The electronic devices described above are used to implement the corresponding methods in any of the foregoing embodiments and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0064] Based on the same inventive concept, corresponding to the methods of any of the above embodiments, this application also provides a non-transitory computer-readable storage medium that stores computer instructions for causing the computer to perform the methods described in any of the above embodiments.

[0065] The computer-readable medium of this embodiment includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0066] The computer instructions stored in the storage medium of the above embodiments are used to cause the computer to perform the methods described in any of the above embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0067] Although this application has been described in conjunction with specific embodiments thereof, many substitutions, modifications, and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may be used with the embodiments discussed.

[0068] The embodiments of this application are intended to cover all such substitutions, modifications, and variations that fall within the broad scope of the claims of this application. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the embodiments of this application should be included within the protection scope of this application.

Claims

1. A method for megakaryocyte recognition based on single-cell transcriptome, characterized in that, include: The transcriptome expression matrix of the single cell to be tested was obtained and quality control and data preprocessing were performed. The expression information of standard genes of megakaryocytes in the preprocessed data was extracted to generate the feature vector of the cell to be tested. Based on the pre-trained machine learning classification model, each cell to be tested is classified as a megakaryocyte, and the prediction result of the cell to be tested is output; wherein, the machine learning classification model is obtained by dividing the expression matrix of single cells labeled as megakaryocytes and non-megakaryocytes into training set and test set according to a preset ratio for training and testing. Based on a pre-defined dual-judgment strategy, the predicted results output by the model are compared with the preset standard gene expression ratio threshold to obtain the predicted megakaryocyte recognition results.

2. The method according to claim 1, characterized in that: Collect single-cell transcriptome data from known megakaryocyte-related organs and developmental stages; wherein the known megakaryocytes are obtained by flow cytometry sorting, published single-cell type annotation, or combination of markers; The single-cell transcriptome data underwent quality control and data preprocessing. Genes that were stably and highly expressed in megakaryocytes across multiple organs and stages and that were distinguishable from non-megakaryocytes were screened to construct a standard gene set for megakaryocytes.

3. The method according to claim 2, characterized in that: The single-cell transcriptome data includes at least yolk sac, peripheral blood, fetal liver / liver, lung, spleen, and bone marrow tissue, and covers data from the embryonic, neonatal, and adult stages.

4. The method according to claim 2, characterized in that: The megakaryocyte standard gene set includes classic megakaryocyte markers Pf4, Itga2b, Gp9, and Itgb3, as well as newly identified or previously underutilized markers Tnik, Dlg2, Pard3b, Ubash3b, Ank3, and Kalrn.

5. The method according to claim 4, characterized in that, The method for constructing the machine learning classification model includes: The single-cell expression matrices labeled as megakaryocytes and non-megakaryocytes are divided into training and testing sets according to a preset ratio. Each cell is composed of a feature vector consisting of the expression level, expression ratio, model score, or a combination thereof of the standard genes of megakaryocytes. A binary classification model is constructed using linear support vector machines, and the model parameters are optimized through cross-validation.

6. The method according to claim 5, characterized in that: One or more of the following metrics were used for model selection: specificity, sensitivity, F1-score, AUC, accuracy, positive predictive value, and negative predictive value.

7. The method according to claim 1, characterized in that, The dual-determination strategy includes: The machine learning classification model outputs the megakaryocyte / non-megakaryocyte category; When the model output is a megakaryocyte and the number or proportion of megakaryocyte standard genes expressed by the cell reaches a preset threshold, it is determined to be a predicted megakaryocyte; wherein, the preset threshold is the expression of no less than 50% of megakaryocyte standard genes.

8. A megakaryocyte recognition system based on single-cell transcriptome, characterized in that, include: The data acquisition and processing module is configured to acquire the transcriptome expression matrix of the single cell to be tested and perform quality control and data preprocessing, extract the expression information of megakaryocyte standard genes in the preprocessed data, and generate the feature vector of the cell to be tested. The model prediction module is configured to classify each cell to be tested as a megakaryocyte by calling a pre-trained machine learning classification model and output the prediction result of the cell to be tested; wherein, the machine learning classification model is obtained by dividing the expression matrix of single cells labeled as megakaryocytes and non-megakaryocytes into training set and test set according to a preset ratio for training and testing. The judgment and recognition module is configured to compare the predicted results output by the model with the preset standard gene expression ratio threshold based on a pre-set dual judgment strategy in order to obtain the predicted megakaryocyte recognition results.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the method as described in any one of claims 1-7.

10. A non-transitory computer-readable storage medium, characterized in that, in, The non-transitory computer-readable storage medium stores computer instructions for causing a computer to perform the method described in any one of claims 1-7.

Images