Machine learning-based gene expression and cell microscopic phenotype association method

Abstract

The present invention provides a machine learning-based gene expression and cell microscopic phenotype association method, which comprises: obtaining cell imaging data and spatial transcriptome data, the cell imaging data comprising a nucleus image of a target cell; performing single cell segmentation based on the cell nucleus image and the spatial transcriptome data through a fine tuning visual field large model to obtain a cell segmentation image of the target cell; determining a cell mask of the target cell based on the cell segmentation image; quantifying the cell mask to obtain the microscopic phenotype of the target cell; determining gene expression of the target cell based on the cell mask and the spatial transcriptome data of the target cell; performing correlation analysis based on the microscopic phenotype of the target cell and the gene expression of the target cell through a preset machine learning model to obtain a correlation result of the microscopic phenotype of the target cell and the gene expression; the association between the microscopic phenotype and the gene expression of the single cell can be realized.

Description

Technical Field

[0001] The present invention relates to the field of cell biotechnology, and in particular to a method for associating gene expression with cell microscopic phenotypes based on machine learning. Background Art

[0002] The series of subtle traits exhibited by plants, whether physical, chemical or biological characteristics, are the result of the expression of specific genes at specific times. These traits are collectively called "micro-phenotypes".

[0003] In the past 20 years, great progress has been made in macroscopic phenotypic traits (organs, plants and canopies), but little attention has been paid to microscopic phenotypic traits and their correlation with macroscopic phenotypic traits. At the same time, gene expression data are usually very complex and contain a lot of information, which makes data interpretation and analysis difficult and requires professional bioinformatics knowledge and skills.

[0004] This shows that the phenotypic trait identification method in the relevant technology has the technical problem of relatively difficult data analysis. Summary of the invention

[0005] The present invention provides a method for associating gene expression and cell microscopic phenotype based on machine learning, which is used to solve the defect of the trait phenotype identification method in the prior art, that is, the data analysis is difficult, and realizes the association between single-cell microscopic phenotype and gene expression.

[0006] The present invention provides a method for associating gene expression and cell microscopic phenotype based on machine learning, comprising the following steps.

[0007] Acquire cell imaging data and spatial transcriptome data, wherein the cell imaging data includes a nucleus image of a target cell; perform single cell segmentation based on the nucleus image and the spatial transcriptome data by fine-tuning a large visual domain model to obtain a cell segmentation image of the target cell; determine a cell mask of the target cell based on the cell segmentation image; quantify the cell mask of the target cell to obtain a microscopic phenotype of the target cell; determine gene expression of the target cell based on the cell mask of the target cell and the spatial transcriptome data; perform association analysis based on the microscopic phenotype of the target cell and the gene expression of the target cell through a preset machine learning model to obtain an association result between the microscopic phenotype of the target cell and the gene expression.

[0008] According to a method for associating gene expression and cell microscopic phenotype based on machine learning provided by the present invention, the method performs single cell segmentation based on the cell imaging data and the spatial transcriptome data by fine-tuning a large model in the visual field to obtain a cell image of the target cell, including: determining the complete image data of the target cell located on the cell imaging data based on the cell nucleus image and the spatial transcriptome data; performing single cell segmentation based on the complete image data by fine-tuning a large model in the visual field to obtain cell boundary positioning; and determining a cell segmentation image of the target cell based on the cell boundary positioning.

[0009] According to a method for associating gene expression and cell microscopic phenotype based on machine learning provided by the present invention, the microscopic phenotype of the target cell includes at least one of the following: cell area, cell perimeter, cell minimum circumscribed circle radius, cell minimum circumscribed ellipse major axis, cell minimum circumscribed ellipse minor axis, cell minimum circumscribed rectangle length, and cell minimum circumscribed rectangle width.

[0010] According to a method for associating gene expression and cell microscopic phenotype based on machine learning provided by the present invention, the gene expression of the target cell is determined based on the cell mask of the target cell and the spatial transcriptome data, including: determining each cell mask position of the target cell; aligning and integrating the spatial transcriptome data based on each cell mask position to obtain the transcriptome data corresponding to each cell mask position; using the transcriptome data corresponding to each cell mask position as the gene expression of each cell mask; and determining the gene expression of the target cell based on the gene expression of each cell mask of the target cell.

[0011] According to a method for associating gene expression and cell micro-phenotype based on machine learning provided by the present invention, after performing an association analysis based on the micro-phenotype of the target cell and the gene expression of the target cell through a preset machine learning model to obtain the association result between the micro-phenotype and gene expression of the target cell, the method further includes: determining a target micro-phenotype of the target cell; based on the association result between the micro-phenotype and gene expression, determining a gene expression whose correlation with the target micro-phenotype is greater than a preset correlation threshold as a candidate gene expression, wherein the candidate gene expression is the target gene expression of the target micro-phenotype.

[0012] According to a method for associating gene expression and cell microscopic phenotype based on machine learning provided by the present invention, after performing an association analysis based on the microscopic phenotype of the target cell and the gene expression of the target cell through a preset machine learning model to obtain the association result between the microscopic phenotype and the gene expression of the target cell, the method further includes: obtaining the single-cell gene expression input by the user; predicting the single-cell gene expression based on the association result between the microscopic phenotype and the gene expression to obtain a predicted microscopic phenotype corresponding to the single-cell gene expression.

[0013] The present invention also provides a device for associating gene expression and cell microscopic phenotype based on machine learning, comprising the following modules: an acquisition module, used to acquire cell imaging data and spatial transcriptome data, wherein the cell imaging data includes a nucleus image of a target cell; a segmentation module, used to perform single-cell segmentation based on the nucleus image and the spatial transcriptome data by fine-tuning a large model of the visual field, to obtain a cell segmentation image of the target cell; a post-segmentation processing module, used to determine a cell mask of the target cell based on the cell segmentation image; a microscopic phenotype quantification module, used to quantify the cell mask of the target cell to obtain the microscopic phenotype of the target cell; a gene expression quantification module, used to determine the gene expression of the target cell based on the cell mask of the target cell and the spatial transcriptome data; an association module, used to perform association analysis based on the microscopic phenotype of the target cell and the gene expression of the target cell through a preset machine learning model, to obtain an association result between the microscopic phenotype of the target cell and the gene expression.

[0014] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for associating gene expression and cell microscopic phenotype based on machine learning as described in any one of the above is implemented.

[0015] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-described methods for associating gene expression and cell microscopic phenotypes based on machine learning.

[0016] The present invention also provides a computer program product, comprising a computer program, which, when executed by a processor, implements any of the above-described methods for associating gene expression and cell microscopic phenotypes based on machine learning.

[0017] The method for associating gene expression and cell microscopic phenotype based on machine learning provided by the present invention can significantly improve the accuracy and efficiency of cell segmentation by fine-tuning a large model in the visual field and performing single cell segmentation using cell nucleus images and spatial transcriptome data. This method of combining multiple data sources can capture the fine structure and boundaries of cells better than traditional methods, thereby obtaining more accurate cell segmentation images. The cell mask determined based on the cell segmentation image and quantified the cell mask can obtain the microscopic performance of the target cell. The gene expression information in the spatial transcriptome data can be accurately matched to a specific cell mask, avoiding the problem of mismatch between gene expression data and cell position in traditional methods. Through a preset machine learning model, the microscopic phenotype of the target cell is associated with gene expression and analyzed to obtain the association result of the microscopic phenotype of the target cell with gene expression, which can reveal the potential connection between the two and help understand how gene expression affects the morphology and function of cells. This solves the technical problem of the difficulty of data analysis in the trait phenotype identification method in the related technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced one by one below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0019] Figure 1 It is a structural schematic diagram of the internal environment and the external environment provided by the present invention.

[0020] Figure 2 It is a schematic flow chart of the method for associating gene expression and cell microscopic phenotype based on machine learning provided by the present invention.

[0021] Figure 3 It is an example diagram of the cell imaging data and spatial transcriptome data provided by the present invention.

[0022] Figure 4 It is a schematic diagram of the structure of the cell boundary positioning provided by the present invention.

[0023] Figure 5 It is a schematic diagram of the integration of cell segmentation images and spatial transcriptome data provided by the present invention.

[0024] Figure 6 It is a schematic diagram of the overall process of gene expression and cell microscopic phenotype based on machine learning provided by the present invention.

[0025] Figure 7It is a schematic diagram of the structure of the device for associating gene expression and cell microscopic phenotype based on machine learning provided by the present invention.

[0026] Figure 8 It is a schematic diagram of the physical structure of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0027] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0028] In the past 20 years, great progress has been made in the phenotypic traits at the macro level (organs, plants and canopies), but little attention has been paid to the phenotypic traits at the micro level and their correlation with the phenotypic traits at the macro level. Phenotyping these traits at the micro level will help to accurately characterize, accurately identify and systematically understand favorable complex traits at the macro level.

[0029] refer to Figure 1 , Figure 1 It is a schematic diagram of the structure of the internal environment and the external environment provided by the present invention, which includes the internal environment, the external environment; the genome, the transcriptome, the proteome, the metabolome, the micro-phenotype and the macro-phenotype.

[0030] The series of subtle traits exhibited by plants, whether physical, chemical or biological characteristics, are the result of the expression of genotypes at specific periods. These traits are collectively referred to as "micro-phenotypes". Plants capture environmental stimuli at different levels, from the canopy to the organs, and show plastic responses. These stimuli from the external environment are converted into stimuli in the internal environment through the exchange and transport of energy and mass at the cellular and tissue levels. The influence of this internal cue is brought about and amplified by signal transduction at the organelle level. Therefore, micro-phenotypes play a key role in bridging the gap between genotypes and complex macroscopic phenotypes, and are very important for the precise design and breeding of crops.

[0031] With the development of molecular biology and cell biology, scientists have begun to focus on microscopic phenotypes at the cellular level, such as cell size, shape, arrangement, and tissue structure, etc. These microscopic phenotypic characteristics can provide more detailed biological information and help to deeply understand the genetic characteristics and physiological mechanisms of crops.

[0032] Machine learning algorithms can learn and extract features from large amounts of data and build association models between gene expression and cell microscopic phenotypes. Through these models, the microscopic trait performance of crops can be predicted more accurately, providing a scientific basis for breeding decisions.

[0033] Based on the spatial transcriptome dataset, relevant algorithms can be used to achieve high-precision cell segmentation and quantify cell microscopic phenotypes. The spatial transcriptome dataset can also be used to obtain gene expression information at corresponding positions, thereby obtaining the gene expression of single cells. Subsequently, certain machine learning algorithms can be used to realize the association study between single-cell microscopic phenotypes and gene expression.

[0034] Optionally, the method for associating gene expression and cell microscopic phenotypes based on machine learning in the embodiment of the present application can be executed by a server, or by a terminal device, or by both a server and a terminal device, taking the example of the method for associating gene expression and cell microscopic phenotypes based on machine learning in this embodiment being executed by a server.

[0035] Figure 2 is a schematic diagram of the process of the method for associating gene expression and cell microscopic phenotype based on machine learning provided by the present invention, such as Figure 2 As shown, the method includes the following steps.

[0036] Step 201 : Acquire cell imaging data and spatial transcriptome data, wherein the cell imaging data includes a nucleus image of a target cell.

[0037] In order to observe the cell nucleus, it is usually necessary to stain the cell nucleus with a specific dye to obtain the stained cell nucleus, and then use a fluorescence microscope or a confocal microscope to image the stained cell nucleus to obtain cell imaging data including the cell nucleus image of the target cell.

[0038] Spatial transcriptomics data (ST data) is a type of biological data that combines transcriptomics and spatial information. The transcriptome refers to the expression of all genes in a cell under specific conditions and time points, that is, the sum of RNA molecules produced by the process of transcribing DNA into RNA.

[0039] Spatial transcriptomics adds spatial information to transcriptomics, that is, it retains the spatial location information of gene expression in tissues. This makes it possible to observe the specific expression location of specific genes in tissues, thereby understanding the relationship between gene expression and tissue structure.

[0040] refer to Figure 3 , Figure 3 It is an example diagram of the cell imaging data and spatial transcriptome data provided by the present invention.

[0041] like Figure 3 As shown, Figure 3 The left picture is a visualization of spatial transcriptome data. Based on the spatial transcriptome data, the gene information and expression information at each coordinate point are determined; Figure 3 The picture on the right is an image of the cell nucleus.

[0042] Continue to refer Figure 2 , step 202, performing single cell segmentation based on the cell nucleus image and the spatial transcriptome data by fine-tuning the large model in the visual field to obtain a cell segmentation image of the target cell.

[0043] In an embodiment of the present invention, based on spatial transcriptome data and imaging data, gene expression information and image information are comprehensively utilized, and high-precision single-cell segmentation is achieved by fine-tuning the large visual field model Segment Anything.

[0044] Acquire the spatial transcriptome data of tissue sections, including gene expression matrix and corresponding spatial location information; preprocess the spatial transcriptome data, including noise removal, normalized expression values, etc.

[0045] Obtain high-resolution nuclear images, including the morphology and position of the nucleus. Perform preprocessing on the nuclear images, such as denoising, contrast enhancement, color correction, etc., to improve the accuracy of segmentation.

[0046] The cell nucleus image is aligned with the spatial transcriptome data so that each spatial transcriptome data point can correspond to the corresponding position in the imaging data.

[0047] Identify a pre-trained large model in the visual field, such as Segment Anything, which has powerful image segmentation capabilities; manually annotate a portion of single cells in the imaging data as a training set, and these annotations should accurately reflect the boundaries of the cells; use the annotated training set to fine-tune the Segment Anything model so that it can more accurately segment single cells; during the fine-tuning process, combine the gene expression information in the spatial transcriptome data as additional feature input to improve the segmentation performance of the model.

[0048] The preprocessed imaging data is input into the fine-tuned Segment Anything model (i.e., the fine-tuned large model in the visual field). The Segment Anything model makes predictions based on the input imaging data and additional feature inputs (i.e., gene expression information in the spatial transcriptome data), and outputs the probability that each pixel belongs to the target cell (i.e., the prediction result). The prediction results of the SegmentAnything model are post-processed, such as threshold segmentation and morphological operations, to obtain a cell segmentation image of the target cell.

[0049] Through the embodiments of the present invention, high-precision single-cell segmentation can be achieved by fine-tuning the large model Segment Anything in the visual field and combining spatial transcriptome data and imaging data.

[0050] Step 203: determine the cell mask of the target cell based on the cell segmentation image.

[0051] In an embodiment of the present invention, the cell segmentation image is converted into a binary image, i.e., a cell mask. For example, a threshold is set, and pixel values ​​above (or below) the threshold are set to white (or non-zero value), and the remaining pixel values ​​are set to black (or zero value).

[0052] In binary images, white areas represent cell areas. Depending on the analysis requirements, specific target cells can be further marked or identified.

[0053] Step 204 , quantifying the cell mask of the target cell to obtain the microscopic phenotype of the target cell.

[0054] In the embodiment of the present invention, OpenCV is used to quantitatively analyze the microscopic phenotype of the segmented cell image.

[0055] Among them, OpenCV (Open Source Computer Vision Library) is an open source computer vision and machine learning software library for processing and analyzing image and video data.

[0056] Step 205 , determining the gene expression of the target cell based on the cell mask and the spatial transcriptome data of the target cell.

[0057] In the embodiment of the present invention, the position of each cell mask can be obtained through the cell segmentation image, and the gene expression information of each position can be understood through the spatial transcriptome data. Therefore, the gene expression information of each cell can be obtained by combining the spatial transcriptome and the cell mask.

[0058] Through the embodiments of the present invention, the gene expression information in the spatial transcriptome data is accurately matched to the specific cell mask, thereby avoiding the problem of mismatch between the gene expression data and the cell position in the traditional method.

[0059] Step 206, using a preset machine learning model, an association analysis is performed based on the microscopic phenotype of the target cells and the gene expression of the target cells to obtain an association result between the microscopic phenotype of the target cells and the gene expression.

[0060] In an embodiment of the present invention, machine learning algorithms such as random forest and support vector machine (SVM) are used to implement the association analysis of single-cell gene expression-micro phenotype, obtain the association results of the micro phenotype and gene expression of the target cell, and then realize the mining of key genes that determine the cell micro phenotype, single-cell micro phenotype prediction analysis, etc.

[0061] Through the above steps of the embodiment of the present invention, cell imaging data and spatial transcriptome data are obtained, wherein the cell imaging data includes a nucleus image of a target cell; single cell segmentation is performed based on the nucleus image and the spatial transcriptome data by fine-tuning the large model of the visual field to obtain a cell segmentation image of the target cell; the cell segmentation image is quantitatively segmented to obtain the microscopic phenotype of the target cell; the cell mask of the target cell is determined based on the cell segmentation image; the gene expression of the target cell is determined based on the cell mask and the spatial transcriptome data of the target cell; through a preset machine learning model, an association analysis is performed based on the microscopic phenotype of the target cell and the gene expression of the target cell to obtain the association result between the microscopic phenotype of the target cell and the gene expression; this helps to understand how gene expression affects the morphology and function of cells; and thereby solves the technical problem that the trait phenotype identification method in the related technology has a high difficulty in data analysis.

[0062] According to a method for associating gene expression and cell microscopic phenotype based on machine learning provided by the present invention, a cell image of a target cell is obtained by fine-tuning a large model in the visual field to perform single cell segmentation based on cell imaging data and spatial transcriptome data, including: Based on the cell nuclear image and spatial transcriptome data, determine the complete image data of the target cell located on the cell imaging data; By fine-tuning the large model in the visual field, single-cell segmentation is performed based on complete image data to obtain cell boundary positioning; Cell segmentation images are used to identify target cells based on cell boundary localization.

[0063] In an embodiment of the present invention, based on the cell nuclear image, the position and morphology of the cell nucleus are identified and located, wherein the cell nuclear image can be obtained by fluorescence microscopy; the spatial transcriptome data is registered with the cell nuclear image to ensure the spatial consistency of the two; the spatial transcriptome data provides detailed information on gene expression in the cell, including the expression level of each gene in each cell or cell group; the cell nuclear position in the cell nuclear image and the gene expression of the spatial transcriptome data are compared to determine the position of the target cell in the spatial transcriptome data. According to the position information of the target cell in the cell nuclear image and the spatial transcriptome data, a complete image containing the target cell is extracted from the cell imaging data.

[0064] The pre-trained deep learning model is fine-tuned using a cell imaging dataset to obtain a fine-tuned large model in the visual field, enabling it to accurately identify and segment individual cells.

[0065] By fine-tuning the large model in the visual field, single-cell segmentation is performed on the complete image data to obtain a segmentation result image. A segmentation mask is performed based on the segmentation result image to obtain the boundary location of each cell. A cell segmentation image is generated based on the boundary location of each cell. Each cell is clearly segmented in the cell segmentation image.

[0066] refer to Figure 4 , Figure 4 It is a schematic diagram of the structure of the cell boundary positioning provided by the present invention.

[0067] like Figure 4 As shown, Figure 4 The left image in the figure is the output of the fine-tuned large model of the visual domain (i.e., the output of the model). Here, the output of the fine-tuned large model of the visual domain is visualized (e.g., the segmentation mask) to obtain the boundary location of each cell (refer to Figure 4 on the right side of the image).

[0068] Through the embodiments of the present invention, the complete image data of the target cell on the cell imaging data is determined based on the cell nuclear image and the spatial transcriptome data, and single cell segmentation is performed by fine-tuning the large model of the visual field, and finally the cell boundary positioning and cell segmentation image are obtained.

[0069] According to a method for associating gene expression and cell microscopic phenotype based on machine learning provided by the present invention, the microscopic phenotype of the target cell includes at least one of the following: cell area, cell perimeter, cell minimum circumscribed circle radius, cell minimum circumscribed ellipse major axis, cell minimum circumscribed ellipse minor axis, cell minimum circumscribed rectangle length, and cell minimum circumscribed rectangle width.

[0070] In an embodiment of the present invention, the microscopic phenotypes of the cell image after quantitative segmentation are obtained by using OpenCV, such as cell area, cell perimeter, cell minimum circumscribed circle radius, cell minimum circumscribed ellipse major axis, cell minimum circumscribed ellipse minor axis, cell minimum circumscribed rectangle length, cell minimum circumscribed rectangle width and other microscopic phenotypes.

[0071] Use OpenCV's cv2.imread() function to read the segmented cell image; use cv2.cvtColor() function to convert the image to a grayscale image for subsequent processing; use cv2.threshold() or an adaptive threshold method (such as cv2.adaptiveThreshold()) to convert the grayscale image to a binary image to more clearly identify the cell boundaries.

[0072] Use the cv2.findContours() function to find the contours of the cells on the binary image; use the cv2.arcLength() function to calculate the perimeter of each contour; use the cv2.minEnclosingCircle() function to find the minimum circumscribed circle of each contour and calculate its radius; use the cv2.fitEllipse() function to fit the minimum circumscribed ellipse of each contour and get the length of its major and minor axes; use the cv2.minAreaRect() function to find the minimum circumscribed rotated rectangle of each contour and get its length and width.

[0073] Use OpenCV drawing functions (such as cv2.circle(), cv2.ellipse(), cv2.rectangle(), etc.) to draw cell outlines and minimum circumscribed shapes on the original image or new image to display the results intuitively.

[0074] Through the embodiments of the present invention, OpenCV can be used to perform quantitative analysis of microscopic phenotypes on segmented cell images.

[0075] According to a method for associating gene expression and cell microscopic phenotype based on machine learning provided by the present invention, the gene expression of target cells is determined based on the cell mask and spatial transcriptome data of the target cells, including: Determine the position of each cell mask of the target cell; Based on the alignment and integration of each cell mask position and the spatial transcriptome data, the transcriptome data corresponding to each cell mask position is obtained; The transcriptome data corresponding to each cell mask position is used as the gene expression of each cell mask; Based on the gene expression of each cell mask of the target cell, the gene expression of the target cell is determined.

[0076] In an embodiment of the present invention, a cell mask image of a target cell corresponding to a cell segmentation image is determined, wherein the region of the target cell is marked as 1 (white) and other regions are marked as 0 (black); the cell mask image is traversed to determine all pixel positions marked as 1 as cell mask positions, which represent the boundary or interior of the target cell.

[0077] Each cell mask position is mapped to the coordinate system of the spatial transcriptome data, and the gene expression value at the corresponding position is extracted to obtain the transcriptome data corresponding to each cell mask position.

[0078] For example, if the spatial transcriptome data is in pixels, the cell mask positions are converted to the corresponding pixel coordinates. If the spatial transcriptome data is in regions (such as small regions in tissue sections), the cell mask is matched to the above regions.

[0079] The extracted (spatial) transcriptome data were associated with the cell mask positions, ensuring that each cell mask had a corresponding gene expression dataset.

[0080] refer to Figure 5 , Figure 5 It is a schematic diagram of the integration of cell segmentation images and spatial transcriptome data provided by the present invention.

[0081] like Figure 5 As shown, Figure 5 The left image in the figure integrates the cell segmentation image with the spatial transcriptome data to extract the transcriptome information at each cell location, for example, for cell 1 (ref. Figure 5 The spatial transcriptome information corresponding to the position of the cell segmentation image is extracted as the gene expression of the cell.

[0082] Through the embodiments of the present invention, the position of each cell mask of the target cell can be determined, and these positions can be aligned and integrated with the spatial transcriptome data to obtain the transcriptome data corresponding to each cell mask, and finally determine the gene expression of the target cell.

[0083] According to a method for associating gene expression and cell microscopic phenotype based on machine learning provided by the present invention, after performing an association analysis based on the microscopic phenotype of target cells and the gene expression of target cells through a preset machine learning model to obtain the association result between the microscopic phenotype and gene expression of the target cells, the method further includes: Determine the target microscopic phenotype of the target cell; Based on the association result between the micro-phenotype and gene expression, a gene expression whose correlation with the target micro-phenotype is greater than a preset correlation threshold is determined as a candidate gene expression, wherein the candidate gene expression is the target gene expression of the target micro-phenotype.

[0084] In the embodiments of the present invention, microscopic phenotype refers to the morphological, structural or functional characteristics exhibited by cells under specific conditions; gene expression refers to the process of gene transcription into mRNA and further translation into proteins, which are the basis of the morphological, structural or functional characteristics exhibited by cells.

[0085] The gene expression data of each cell is matched with the corresponding microscopic phenotype data, and the association between gene expression and microscopic phenotype is analyzed using statistical methods (such as correlation analysis, regression analysis, machine learning algorithms, etc.); based on the results of the association analysis, the gene expression that is significantly correlated with the microscopic phenotype of the target cell is screened out as the target gene expression, where the target gene expression is directly involved in regulating the microscopic phenotype of the target cell.

[0086] According to a method for associating gene expression and cell microscopic phenotype based on machine learning provided by the present invention, after performing an association analysis based on the microscopic phenotype of target cells and the gene expression of target cells through a preset machine learning model to obtain the association result between the microscopic phenotype and gene expression of the target cells, the method further includes: Get single-cell gene expression input by the user; Based on the correlation results between microscopic phenotype and gene expression, single-cell gene expression is predicted to obtain the predicted microscopic phenotype corresponding to the single-cell gene expression.

[0087] In the embodiment of the present invention, the single-cell microscopic phenotype input by the user includes at least one of the cell's morphology, size, structural characteristics, and functional state.

[0088] Collect known single-cell gene expression data and corresponding microscopic phenotype data as training sets, use machine learning algorithms (such as support vector machines, random forests, neural networks, etc.) or deep learning models (such as convolutional neural networks, recurrent neural networks, etc.), use gene expression features as input and microscopic phenotypes as output to train models. Use the trained model to predict the single-cell gene expression input by the user to obtain the predicted microscopic phenotype.

[0089] Through the embodiments of the present invention, by obtaining the single-cell microscopic phenotype input by the user and making predictions based on the correlation results between the microscopic phenotype and gene expression, we can obtain the predicted microscopic phenotype corresponding to the single-cell gene expression, which helps us to understand the functions and characteristics of cells more deeply.

[0090] refer to Figure 6 , Figure 6 It is a schematic diagram of the overall process of gene expression and cell microscopic phenotype based on machine learning provided by the present invention, which specifically includes the following steps.

[0091] High-precision single-cell segmentation: Based on spatial transcriptome data and imaging data, gene expression information and image information are comprehensively utilized to achieve high-precision single-cell segmentation by fine-tuning the large visual field model Segment Anything.

[0092] Single-cell microscopic phenotype quantification: Use OpenCV to quantify the microscopic phenotypes of the segmented cell images, such as cell area, cell perimeter, cell minimum circumscribed circle radius, cell minimum circumscribed ellipse major axis, cell minimum circumscribed ellipse minor axis, cell minimum circumscribed rectangle length, cell minimum circumscribed rectangle width and other microscopic phenotypes.

[0093] Single-cell gene expression quantification: The location of each cell mask can be obtained through the segmented cell image, and the gene expression information at each location can be obtained through the spatial transcriptome data. Therefore, by combining the spatial transcriptome and cell mask images, the gene expression information of each cell can be obtained.

[0094] Here, single-cell gene expression quantification was determined based on the location of a cell membrane stain (CellMask) and spatial transcriptome data.

[0095] Gene expression-micro phenotype association analysis: Use machine learning algorithms such as random forest and SVM to realize single-cell gene expression-micro phenotype association analysis, to mine key genes that determine cell micro phenotypes, and to predict and analyze single-cell micro phenotypes.

[0096] The device for associating gene expression and cell microscopic phenotype based on machine learning provided by the present invention is described below. The device for associating gene expression and cell microscopic phenotype based on machine learning described below and the method for associating gene expression and cell microscopic phenotype based on machine learning described above can be referenced to each other.

[0097] refer to Figure 7 , Figure 7 It is a schematic diagram of the structure of the device for associating gene expression and cell microscopic phenotype based on machine learning provided by the present invention, which includes: an acquisition module 701, a segmentation module 702, a segmentation post-processing module 703, a microscopic phenotype quantification module 704, a gene expression quantification module 705 and an association module 706.

[0098] An acquisition module 701 is used to acquire cell imaging data and spatial transcriptome data, wherein the cell imaging data includes a nucleus image of a target cell; The segmentation module 702 is used to perform single cell segmentation based on the cell nucleus image and the spatial transcriptome data by fine-tuning the large model of the visual field to obtain a cell segmentation image of the target cell; A post-segmentation processing module 703 is used to determine a cell mask of a target cell based on the cell segmentation image; A microscopic phenotype quantification module 704 is used to quantify the cell mask of the cell segmentation image to obtain the microscopic phenotype of the target cell; A gene expression quantification module 705, for determining the gene expression of the target cell based on the cell mask and the spatial transcriptome data of the target cell; The association module 706 is used to perform association analysis based on the microscopic phenotype of the target cell and the gene expression of the target cell through a preset machine learning model to obtain the association result between the microscopic phenotype of the target cell and the gene expression.

[0099] Specifically, the above-mentioned device for associating gene expression and cell microscopic phenotype based on machine learning provided by the present invention can implement all the method steps implemented by the above-mentioned method embodiment for associating gene expression and cell microscopic phenotype based on machine learning, and can achieve the same technical effect. The parts and beneficial effects of this embodiment that are the same as those of the method embodiment will not be described in detail here.

[0100] Figure 8 is a schematic diagram of the physical structure of the electronic device provided by the present invention, such as Figure 8 As shown, the electronic device may include: a processor (processor) 810, a communication interface (Communications Interface) 820, a memory (memory) 830 and a communication bus 840, wherein the processor 810, the communication interface 820, and the memory 830 communicate with each other through the communication bus 840. The processor 810 can call the logic instructions in the memory 830 to execute a method for associating gene expression and cell microscopic phenotype based on machine learning, the method comprising: acquiring cell imaging data and spatial transcriptome data, wherein the cell imaging data includes a nucleus image of a target cell; performing single cell segmentation based on the nucleus image and the spatial transcriptome data by fine-tuning the large model of the visual field to obtain a cell segmentation image of the target cell; performing quantitative segmentation on the cell segmentation image to obtain a microscopic phenotype of the target cell; determining a cell mask of the target cell based on the cell segmentation image; performing quantitative segmentation on the cell mask cell segmentation image of the target cell to obtain a microscopic phenotype of the target cell; determining the gene expression of the target cell based on the cell mask and the spatial transcriptome data of the target cell; performing association analysis based on the microscopic phenotype of the target cell and the gene expression of the target cell through a preset machine learning model to obtain an association result between the microscopic phenotype of the target cell and the gene expression.

[0101] In addition, the logic instructions in the above-mentioned memory 830 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0102] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the machine learning-based gene expression and cell microscopic phenotype association method provided by the above methods, the method including: acquiring cell imaging data and spatial transcriptome data, wherein the cell imaging data includes a cell nucleus image of a target cell; performing single cell segmentation based on the cell nucleus image and the spatial transcriptome data by fine-tuning a large model in the visual field to obtain a cell segmentation image of the target cell; performing quantitative segmentation on the cell segmentation image to obtain a microscopic phenotype of the target cell; determining a cell mask of the target cell based on the cell segmentation image; performing quantitative segmentation on the cell mask cell segmentation image of the target cell to obtain a microscopic phenotype of the target cell; determining the gene expression of the target cell based on the cell mask and the spatial transcriptome data of the target cell; performing association analysis based on the microscopic phenotype of the target cell and the gene expression of the target cell through a preset machine learning model to obtain an association result between the microscopic phenotype of the target cell and the gene expression of the target cell.

[0103] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the machine learning-based gene expression and cell microscopic phenotype association method provided by the above-mentioned methods, the method comprising: acquiring cell imaging data and spatial transcriptome data, wherein the cell imaging data includes a cell nucleus image of a target cell; performing single cell segmentation based on the cell nucleus image and the spatial transcriptome data by fine-tuning a large visual field model to obtain a cell segmentation image of the target cell; performing quantitative segmentation on the cell segmentation image to obtain a microscopic phenotype of the target cell; determining a cell mask of the target cell based on the cell segmentation image; performing quantitative segmentation on the cell mask cell segmentation image of the target cell to obtain a microscopic phenotype of the target cell; determining the gene expression of the target cell based on the cell mask and the spatial transcriptome data of the target cell; performing association analysis based on the microscopic phenotype of the target cell and the gene expression of the target cell through a preset machine learning model to obtain an association result between the microscopic phenotype of the target cell and the gene expression of the target cell.

[0104] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, i.e., they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Those of ordinary skill in the art may understand and implement it without creative effort.

[0105] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0106] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for associating gene expression and cell microscopic phenotype based on machine learning, characterized in that: include: Acquiring cell imaging data and spatial transcriptome data, wherein the cell imaging data includes a cell nucleus image of a target cell; Performing single cell segmentation based on the cell nucleus image and the spatial transcriptome data by fine-tuning the large model in the visual field to obtain a cell segmentation image of the target cell; Determine a cell mask of the target cell based on the cell segmentation image; quantifying the cell mask of the target cell to obtain the microscopic phenotype of the target cell; Determining gene expression of the target cell based on the cell mask of the target cell and the spatial transcriptome data; Through a preset machine learning model, an association analysis is performed based on the microscopic phenotype of the target cells and the gene expression of the target cells to obtain the association result between the microscopic phenotype of the target cells and the gene expression.

2. The method for associating gene expression and cell microscopic phenotype based on machine learning according to claim 1, characterized in that: The method of performing single cell segmentation based on the cell imaging data and the spatial transcriptome data by fine-tuning the large model of the visual field to obtain the cell image of the target cell includes: Based on the cell nuclear image and the spatial transcriptome data, determining the complete image data of the target cell located on the cell imaging data; Performing single cell segmentation based on the complete image data by fine-tuning the large model of the visual field to obtain the cell boundary location; A cell segmentation image of the target cell is determined based on the cell boundary positioning.

3. The method for associating gene expression and cell microscopic phenotype based on machine learning according to claim 1, characterized in that: The microscopic phenotype of the target cell includes at least one of the following: cell area, cell perimeter, cell minimum circumscribed circle radius, cell minimum circumscribed ellipse major axis, cell minimum circumscribed ellipse minor axis, cell minimum circumscribed rectangle length, and cell minimum circumscribed rectangle width.

4. The method for associating gene expression and cell microscopic phenotype based on machine learning according to claim 1, characterized in that: The determining the gene expression of the target cell based on the cell mask of the target cell and the spatial transcriptome data comprises: determining each cell mask position of the target cell; Based on the alignment and integration of each cell mask position and the spatial transcriptome data, the transcriptome data corresponding to each cell mask position is obtained; Using the transcriptome data corresponding to each cell mask position as the gene expression of each cell mask; Based on the gene expression of each cell mask of the target cell, the gene expression of the target cell is determined.

5. The method for associating gene expression and cell microscopic phenotype based on machine learning according to claim 1, characterized in that: After performing association analysis based on the microscopic phenotype of the target cell and the gene expression of the target cell through a preset machine learning model to obtain the association result between the microscopic phenotype and the gene expression of the target cell, the method further includes: Determine the target microscopic phenotype of the target cell; Based on the association result between the micro-phenotype and gene expression, a gene expression whose correlation with the target micro-phenotype is greater than a preset correlation threshold is determined as a candidate gene expression, wherein the candidate gene expression is the target gene expression of the target micro-phenotype.

6. The method for associating gene expression and cell microscopic phenotype based on machine learning according to claim 1, characterized in that: After performing association analysis based on the microscopic phenotype of the target cell and the gene expression of the target cell through a preset machine learning model to obtain the association result between the microscopic phenotype and the gene expression of the target cell, the method further includes: Get single-cell gene expression input by the user; The single cell gene expression is predicted based on the correlation result between the microscopic phenotype and the gene expression to obtain a predicted microscopic phenotype corresponding to the single cell gene expression.

7. A device for associating gene expression and cell microscopic phenotype based on machine learning, characterized in that: include: An acquisition module, used to acquire cell imaging data and spatial transcriptome data, wherein the cell imaging data includes a cell nucleus image of a target cell; A segmentation module, used for performing single cell segmentation based on the cell nucleus image and the spatial transcriptome data by fine-tuning a large model of the visual field to obtain a cell segmentation image of the target cell; A post-segmentation processing module, used for determining a cell mask of the target cell based on the cell segmentation image; A microscopic phenotype quantification module, used for quantifying the cell mask of the target cell to obtain the microscopic phenotype of the target cell; A gene expression quantification module, for determining the gene expression of the target cell based on the cell mask of the target cell and the spatial transcriptome data; The association module is used to perform association analysis based on the microscopic phenotype of the target cell and the gene expression of the target cell through a preset machine learning model to obtain the association result between the microscopic phenotype of the target cell and the gene expression.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the method for associating gene expression and cell microscopic phenotype based on machine learning as described in any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for associating gene expression and cell microscopic phenotype based on machine learning as described in any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method for associating gene expression and cell microscopic phenotype based on machine learning as described in any one of claims 1 to 6 is implemented.

Images