Systems and methods of single-cell spatial multiomics and metabolomics analyses

Abstract

A system may receive a plurality of images that capture at least one cell having a complex shape, the plurality of images comprising first and second stained images. A system may receive a data matrix that represents a gene expression profile of the at least one cell. A system may combine the first and second stained images to create an overlay image. A system may analyze at least one of the second stained image or the overlay image to delineate a boundary of the at least one cell. A system may create a mask based on the boundary of the at least one cell in the at least one of the second stained image or the overlay image. A system may combine the mask and the data matrix to form a combined image for the at least one cell.

Description

SYSTEMS AND METHODS OF SINGLE-CELL SPATIAL MULTIOMICS AND METABOLOMICS ANALYSESCROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims the benefit of U.S. provisional patent application No. 63 / 564,730, filed on March 13, 2024, and titled “SYSTEMS AND METHODS OF SINGLECELL SPATIAL MULTIOMICS AND METABOLOMICS ANALYSES,” the disclosure of which is expressly incorporated herein by reference in its entirety.STATEMENT REGARDING GOVERNMENT SUPPORT

[0002] This invention was made with government support under NS 133881 awarded by the National Institutes of Health. The government has certain rights in the invention.BACKGROUND

[0003] Single-cell omics analysis includes the comprehensive characterization and quantification of the genomes, transcriptomes, proteomes, epigenomes, and metabolomes of single cells, thereby enabling a deeper understanding of cellular diversity, function, and dynamics within heterogeneous tissues and populations. For example, single-cell omics analysis can be used to evaluate drug efficacy and revolutionize disease diagnosis and prognosis because it can profile 36,000 genes (instead of just evaluating a few “marker” genes). Improvements to single-cell omics can improve medical research.SUMMARY

[0004] In some aspects, the techniques described herein relate to a computer- implemented method of single-cell spatial omics analysis including: receiving a plurality of images that capture at least one cell having a complex shape, the plurality of images including a first stained image and a second stained image, wherein the first stained image enhances the visibility or contrast of the at least one cell's nucleus, and wherein the second stained image enhances the visibility or contrast of the at least one cell's edges; receiving a data matrix that represents a gene expression profile of the at least one cell; combining the first stained image and the second stained image to create an overlay image; analyzing at least one of the second stained image or the overlay image to delineate a boundary of the at least one cell; creating a mask based on the boundary of the at least one cell in the at least one of the second stained image or theoverlay image; and combining the mask and the data matrix to form a combined image for the at least one cell.

[0005] In some aspects, the method further includes performing an omics analysis of the combined image.

[0006] In some aspects, the omics analysis includes determining transcript counts for the at least one cell.

[0007] In some aspects, the omics analysis includes determining a spatial distribution of the gene expression profile of the at least one cell.

[0008] In some aspects, the omics analysis is at least one of a cell typing analysis, a cell-cell communication analysis, an analysis of the number and distribution of different cell types, a differential gene expression analysis, a marker gene identification analysis, a proteinprotein interaction network analysis, a cell trajectory analysis, or a functional enrichment analysis.

[0009] In some aspects, the step of analyzing at least one of the second stained image or the overlay image to delineate the boundary of the at least one cell includes: filtering the at least one of the second stained image or the overlay image based on at least one of a signal intensity or a structure size; performing an edge detection process to identify a plurality of pixels that form the boundary of the at least one cell; retaining the plurality of pixels that form the boundary of the at least one cell, wherein the boundary of the at least one cell includes a plurality of gaps between pixels of the plurality of pixels; and connecting the boundary of the at least one cell by extending through the plurality of gaps.

[0010] In some aspects, the method further includes receiving a mass spectrometry image that captures the at least one cell, wherein the mass spectrometry image is a spatial representation of a plurality of lipids or metabolites within the at least one cell; combining information associated with the combined image and the mass spectrometry image to form an integrated image; and analyzing the integrated image to determine a metabolic profile of the at least one cell.

[0011] In some aspects, the information associated with the combined image is the boundary of the at least one cell and a type of the at least one cell.

[0012] In some aspects, the method further includes determining a signal intensity and a location of at least one metabolite and a metabolic state of the at least one cell.

[0013] In some aspects, the at least one cell is elongated, cylindrical, polygonal spindle shaped, pyramid shaped, dome shaped, or star shaped.

[0014] In some aspects, the at least one cell includes a plurality of branching structures.

[0015] In some aspects, the complex shape is not a round or elliptical shape.

[0016] In some aspects, the at least one cell is a neuronal cell, an astrocyte, a fibroblast, a mesenchymal stem cell, a smooth muscle cell, a skeletal muscle fiber, an osteoclast, a keratinocyte, an endothelial cell, or a Sertoli cell.

[0017] In some aspects, the at least one cell is stained with a dye that binds to deoxyribonucleic acid (DNA) in the first stained image.

[0018] In some aspects, the at least one cell is stained with an antibody in the second stained image.

[0019] In some aspects, the techniques described herein relate to a method including: performing, for a biological sample from a subject, the method of single-cell spatial omics analysis as described above; and determining a diagnosis, prognosis, or treatment plan for the subject based on one or more respective cell-by-gene matrices for one or more cells of the biological sample.

[0020] In some aspects, the method further includes administering a treatment to the subject based on the diagnosis, prognosis, or treatment plan.

[0021] In some aspects, the techniques described herein relate to a computer- implemented method including: receiving a plurality of images that capture at least one cell having a complex shape, the plurality of images including a first stained image and a second stained image, wherein the first stained image enhances the visibility or contrast of the at least one cell's nucleus, and wherein the second stained image enhances the visibility or contrast of the at least one cell's edges; combining the first stained image and the second stained image to create an overlay image; analyzing at least one of the second stained image or the overlay image to delineate a boundary of the at least one cell; and segmenting the at least one cell based on the boundary of the at least one cell in the at least one of the second stained image or the overlay image.

[0022] In some aspects, the techniques described herein relate to a computer- implemented method of single-cell spatial omics analysis including: receiving a plurality of images that capture at least one cell, the plurality of images including a first stained image and a second stained image, wherein the first stained image enhances the visibility or contrast of the at least one cell's nucleus, and wherein the second stained image enhances the visibility or contrast of the at least one cell's edges; receiving a data matrix that represents a gene expression profile ofthe at least one cell; combining the first stained image and the second stained image to create an overlay image; analyzing at least one of the second stained image or the overlay image to delineate a boundary of the at least one cell; creating a mask based on the boundary of the at least one cell in the at least one of the second stained image or the overlay image; and combining the mask and the data matrix to form a combined image for the at least one cell. Additionally, the step of analyzing at least one of the second stained image or the overlay image to delineate the boundary of the at least one cell includes: filtering the at least one of the second stained image or the overlay image based on at least one of a signal intensity or a structure size; performing an edge detection process to identify a plurality of pixels that form the boundary of the at least one cell; retaining the plurality of pixels that form the boundary of the at least one cell, wherein the boundary of the at least one cell includes a plurality of gaps between pixels of the plurality of pixels; and connecting the boundary of the at least one cell by extending through the plurality of gaps.

[0023] In some aspects, the techniques described herein relate to a computer- implemented method of single-cell spatial metabolomics analysis including: receiving a transcriptomic image for at least one cell, wherein the transcriptomic image is a spatial representation of a gene expression profile of the at least one cell; receiving a mass spectrometry image that captures the at least one cell, wherein the mass spectrometry image is a spatial representation of a plurality of lipids or metabolites within the at least one cell; combining information associated with the transcriptomic image and the mass spectrometry image to form an integrated image; and analyzing the integrated image to determine a metabolic profile of the at least one cell.

[0024] In some aspects, the information associated with the transcriptomic image is a boundary of the at least one cell and a type of the at least one cell.

[0025] In some aspects, the method further includes determining a signal intensity and a location of at least one metabolite and a metabolic state of the at least one cell.

[0026] In some aspects, the method further includes generating display data for the integrated image.

[0027] In some aspects, the transcriptomic image for the at least one cell is obtained by performing a single-cell spatial omics analysis.

[0028] In some aspects, the method includes: performing, for a biological sample from a subject, the method of single-cell spatial metabolomics analysis as described above; anddetermining a diagnosis, prognosis, or treatment plan for the subject based on one or more respective metabolic profiles for one or more cells of the biological sample.

[0029] In some aspects, the method further includes administering a treatment to the subject based on the diagnosis, prognosis, or treatment plan.

[0030] In some aspects, the techniques described herein relate to a method including: seeding a plurality of cells on a surface; treating the plurality of cells with one or more therapeutic agents; performing, for a biological sample from the plurality of cells, the method of single-cell spatial omics analysis as described above; and determining a therapeutic effect of the one or more therapeutic agents based on the single-cell spatial omics analysis.

[0031] In some aspects, the techniques described herein relate to a method including: seeding a plurality of cells on a surface; treating the plurality of cells with one or more therapeutic agents; performing, for a biological sample from the plurality of cells, the method of single-cell spatial metabolomics analysis as described above; and determining a therapeutic effect of the one or more therapeutic agents based on the single-cell spatial metabolomics analysis.

[0032] It should be understood that the above-described subject matter may also be implemented as a computer-controlled apparatus, a computer process, a computing system, or an article of manufacture, such as a computer-readable storage medium.

[0033] Other systems, methods, features and / or advantages will be or may become apparent to one with skill in the art upon examination of the following drawings and detailed description. It is intended that all such additional systems, methods, features and / or advantages be included within this description and be protected by the accompanying claims.BRIEF DESCRIPTION OF THE DRAWINGS

[0034] The components in the drawings are not necessarily to scale relative to each other. Like reference numerals designate corresponding parts throughout the several views.

[0035] FIGURE 1 is a flowchart of an example method for single-cell spatial omics analysis according to an implementation described herein.

[0036] FIGURE 2 is a flowchart of an example method for cell segmentation according to an implementation described herein.

[0037] FIGURE 3 is a flowchart of an example method for single-cell spatial metabolomics analysis according to an implementation described herein.

[0038] FIGURE 4 is an example computing device.

[0039] FIGURE 5 illustrates various images of an astrocyte (i.e. a cell) according to an implementation described herein.

[0040] FIGURE 6 illustrates example techniques and platforms for performing singlecell transcriptomics according to an implementation described herein.

[0041] FIGURE 7 is a flowchart illustrating a process for analyzing a stained image to delineate a boundary of a cell according to an implementation described herein.

[0042] FIGURE 8 illustrates results comparing the method described with regard to Fig. 1 with a single-cell spatial omics technique using a conventional cell segmentation method (e.g. Cellpose).

[0043] FIGURE 9 illustrates example mass spectroscopy and integrated images according to an implementation described herein.

[0044] FIGURE 10 illustrates an evaluation of the lipids in different cell types using the method described with regard to Figure 3.

[0045] FIGURE 11 is an example overlay image used to delineate cell boundaries according to an implementation described herein.

[0046] FIGURE 12 is an example cell-by-gene matrix according to an implementation described herein.

[0047] FIGURE 13(A-D) shows that TransMetaSegmentation (TMS) enables singlecell spatial metabolomics. Figure 13A shows the 'open-book' mounting technique enables the same cells to be analyzed using two different technologies. Figure 13B shows the Xenium workflow consists of probe hybridization, ligation, rolling circle amplification, imaging, and data analysis. Figure 13C shows the MALDI MSI procedure involves the application of conductive reagents, mass spectrometry imaging, and subsequent data analysis. Figure 13D shows that single-cell spatial metabolomics can be achieved by integrating the spatial distribution of metabolites from MALDI MSI with cell boundary and cell type information from Xenium.

[0048] FIGURE 14(A-J) shows how Xenium profiles the cell types and boundaries. Figure 14(A-C) shows that cell boundaries are defined by Cellpose. Figure 14D shows a tSNE showing the different clusters of brain cells based on transcriptomic analysis. Figure 14E shows the spatial distribution of the different clusters. Figure 14F shows a cell-by-gene matrix of the brain cells. Figure 14G shows the percentages of different clusters in the 122,640 cells annotated in the brain section. (H-J) Defining cell types based on gene expression patterns. PVM, perivascular macrophages; VLMS, vascular leptomeningeal cells; cl, cluster 1.

[0049] FIGURE 15(A-L) shows the spatial distribution of putatively identified lipids, including phospholipids, phosphatidic acids, and sphingolipids. Figure 15(A-D) shows spatial distribution of phosphatidylethanolamines (PE), e.g., [PE(O-36:3)-H]‘ (m / z= 726.5436, as inFigure 15A), |PE(O-38:3)-H|’ (m / z=754.5758, as in Figure 15B), |PE(O-38:7)-H|’ (m / z=746.5126, as in Figure 15C), and [PE(38:1)-H]‘ (m / z=772.5903, as in Figure 15D). Figure 15E shows the spatial distribution of phosphatidic acid (PA) [PA(36:1)-H]‘ (m / z=701.5122).Figure 15(F-G) shows the spatial distribution of hexosylceramides (HexCer), e.g., [HexCer(42:2;O3)-H]’ (m / z=824.6632, as in Figure 15F) and [HexCer(42:2;O2)-H]’ (m / z=808.669, as in Figure 15G). Figure 15(H-K) shows the spatial distribution of sulfated hexosylceramides (SHexCer), e.g., [SHexCer(36: l;O2)-H]‘ (m / z=806.5486, as in Figure 15H), [SHexCer(36:l;O3)-H]- (m / z=822.5425, as in Figure 151), [SHexCer(40:2;O2)-H]’ (m / z=860.5971, as in Figure 15J), and [SHexCer(38:l ;O3)-H]‘ (m / z=850.5723, as in Figure 15K). Figure 15L shows the spatial distribution of ceramide phosphoethanolamines (CerPE) [CerPE(44:2;O2)-H]‘ (m / z=797.6538, as in Figure 15L). Scale bars are 800 pm.

[0050] FIGURE 16(A-E) shows an affine transformation for co-registration of images, followed by quantification of MSI signal. Figure 16A shows a stereo-seq image of a mouse brain section. Figure 16B shows a magnified view of Figure 16A, showing cell boundaries. Each color represents one unique cell type. Figure 16C shows that cell boundaries were identified and labeled with cell type information. The image background was removed to create a transparent background. Figure 16D shows MSI images recreated using the CSV file from SCiLS software. Figure 16E shows overlaying the MSI image with a Stereo-seq image allowed for quantifying metabolite signal intensity in individual cells (See Method).

[0051] FIGURE 17(A-L) shows that TMS enables the evaluation of the lipids in different cell types. Figure 17(A-B) shows the overlay of an image from MSI with a cell segmentation image from single-cell spatial transcriptomics, yielding a single-cell spatial lipid image. The spatial distribution of [SHexCer(42:l ;O3)-H]‘ (m / z=906.6349) in the control mouse brain (as in Figure 17 A) and Canavan mouse brain (as in Figure 17B). Figure 17(C-D) shows the spatial distribution of [SHexCer(42:2;O2)-H]‘ (m / z=888.6289) in the control mouse brain (as in Figure 17C) and Canavan mouse brain (as in Figure 17D). Figure 17(E-F) shows the quantification of [SHexCer(42: 1 ;O3)-H]‘ (m / z=906.6349) in oligodendrocytes (as in Figure 17E) and astrocytes(as in Figure 17F). Figure 17(G-H) shows the quantification of [SHexCer(42:2;O2)-H]_(m / z=888.6289) in oligodendrocytes (as in Figure 17G) and astrocytes (as in Figure 17H). Figure 17(I-J) The quantification of [PE(38:5)-H]' (m / z=764.5236) in astrocytes (as in Figure 171) and endothelial cells (as in Figure 17J). Figure 17(K-L) shows the quantification of [PI(40:6)-H]‘ (m / z=909.5515) in oligodendrocytes (as in Figure 17K) and neurons (as in Figure 17L). As, astrocyte; O, oligodendrocyte; En, endothelial cell; M, microglia; N, neuron; OPC, oligodendrocyte progenitor cells; P, pericyte; V, vascular leptomeningeal cells. Scale bars are 800 pm. Values are mean ± SD. p values determined by Mann- Whitney U tests. **p < 0.05. n.s., not significant. Lipids were putatively identified using the LIPID MAPS database.

[0052] FIGURE 18(A-J) shows differentially expressed genes identified by Xenium Spatial distribution of differentially expressed genes Clmn (as in Figure 18A), Car4 (as in Figure 18B), Car4 (as in Figure 18C), Meis2 (as in Figure 18D), Nrnl (as in Figure 18E), Nr2f2 (as in Figure 18F), Necabl (as in Figure 18G), Pvalb (as in Figure 18H), Rorb (as in Figure 181), and Slcl7a7 (as in Figure 18J).

[0053] FIGURE 19(A-K) shows a rigid two-point approach to achieve single-cell spatial metabolomics. Figure 19A shows how to open the single-cell spatial transcriptomic image in Xenium Explorer 3, select “Images” and “Cells,” and save the image. Figure 19B shows how to open the MALDI MSI image in SCiLS Lab Version 2024b. Figure 19C shows how to import the transcriptomic image into SCiLS Lab. (D-E) Set 2 anchor points for the transcriptomic image. Figure 19F shows how to set 2 anchor points for the MALDI MSI image. Figure 19G shows how to align the transcriptomic image with the MALDI MSI image using the anchor points. Figure 19H shows how to adjust the transparency of the images to see the MALDI MSI image through the transcriptomic image. Figure 191 shows how to draw shapes on top of the cell boundaries. Figure I9J shows the quantification of the metabolite signals in the shapes or cells. Figure 19K shows the statistical analysis of the metabolite signal.DETAILED DESCRIPTION

[0054] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art. Methods and materials similar or equivalent to those described herein can be used in the practice or testing of the present disclosure. As used in the specification, and in the appended claims, the singular forms “a,” “an,” “the” include plural referents unless the context clearly dictates otherwise. Theterm “comprising” and variations thereof as used herein is used synonymously with the term “including” and variations thereof and are open, non-limiting terms. The terms “optional” or “optionally” used herein mean that the subsequently described feature, event or circumstance may or may not occur, and that the description includes instances where said feature, event or circumstance occurs and instances where it does not. Ranges may be expressed herein as from "about" one particular value, and / or to "about" another particular value. When such a range is expressed, an aspect includes from the one particular value and / or to the other particular value. Similarly, when values are expressed as approximations, by use of the antecedent "about," it will be understood that the particular value forms another aspect. It will be further understood that the endpoints of each of the ranges are significant both in relation to the other endpoint, and independently of the other endpoint.

[0055] As used herein, the terms "about" or "approximately" when referring to a measurable value such as an amount, a percentage, and the like, is meant to encompass variations of ±20%, ±10%, ±5%, or ±1% from the measurable value.

[0056] “Administration” of “administering” to a subject includes any route of introducing or delivering to a subject an agent. Administration can be carried out by any suitable means for delivering the agent. Administration includes self-administration and the administration by another.

[0057] The term “subject” is defined herein to include animals such as mammals, including, but not limited to, primates (e.g., humans), cows, sheep, goats, horses, dogs, cats, rabbits, rats, mice and the like. In some embodiments, the subject is a human.

[0058] Implementations of the present disclosure include improvements to single-cell omics. Existing methodologies face challenges in scalability, sensitivity, and specificity, often requiring intricate sample preparation and sophisticated analytical tools. Implementations of the present disclosure can efficiently and accurately segment and analyze single cells with sub- cellular resolution in a high-throughput manner, minimizing sample loss and artifacts while maximizing the recovery of biological information.

[0059] Example Methods of Single- Cell Spatial Omics

[0060] Referring now to Fig. 1 , a flowchart illustrating an example method for singlecell spatial omics analysis is shown. This disclosure contemplates that the logical operations of Fig. 1 can be performed by a computing device. An example computing device is shown in Fig. 4.

[0061] At step 110, the method includes receiving a plurality of images that capture at least one cell having a complex shape. The images may capture a single cell or a plurality of cells. Additionally, the at least one cell may be part of a tissue sample. It should be understood that a tissue sample is provided only as an example. Alternatively, the at least one cell may be part of a cell culture, a cell suspension, or other group of cells. As noted above, the at least one cell has a complex shape. Existing automated cell segmentation techniques (e.g. Cellpose, which is an open-source anatomical segmentation algorithm) are unable to accurately segment cells having complex shapes as described in the Examples below. For example, such existing automated cell segmentation techniques tend to oversimplify cellular shapes, categorizing cells as either round or elliptical. This negatively impacts spatial cell omics analyses as described in the Examples below. Accordingly, as used herein, a complex shape is not a round or elliptical shape. Example complex shapes include, but are not limited to, elongated, cylindrical, polygonal spindle, pyramid, dome, or star shapes. Example cells having complex shapes include, but are not limited to, a neuronal cell (multipolar or bipolar with elongated axons and dendrites), an astrocyte (star-shaped), a fibroblast, a mesenchymal stem cell (spindle-like shape), a smooth muscle cell (spindle-shaped), a skeletal muscle fiber (cylindrical and multinucleated), an osteoclasts (dome-shaped and multinucleated), a keratinocyte (polygonal), an endothelial cell (elongated polygons, usually hexagonal), or a Sertoli cell (pyramid-shaped). In the Examples below, the at least one cell is a neuronal cell. It should be understood that a neuronal cell is provided only as an example cell having a complex shape.

[0062] The plurality of images comprises a first stained image and a second stained image. The first stained image and the second stained image capture the at least one cell having a complex shape. In other words, the at least one cell captured by the first and second stained images is the same cell, which has a complex shape. For example, the at least one cell may be part of a tissue sample that is fixed to a substrate. The tissue sample is then stained prior to imaging. Cell staining is a technique that involves the application of dyes or other markers to visualize specific cellular structures, specific proteins, or other biomolecules within cells or tissues. Staining is used to enhance the contrast and visibility of cells, allowing researchers to study their morphology, structure, and function.

[0063] An example cell 502 having a complex shape (an astrocyte) is shown in Fig. 5. It should be understood that an astrocyte is only provided as an example of a cell having a complex shape. This disclosure contemplates using the method of Fig. 1 on other cells. The first stained image enhances the visibility or contrast of the at least one cell's nucleus. In other words,the tissue sample is stained with a dye that binds to genetic materials (e.g., deoxyribonucleic acid (DNA)) found in the at least one cell’s nucleus, and the first stained image is acquired. DAPI (4',6-diamidino-2-phenylindole) is an example dye that enhances the visibility or contrast of the nucleus. It should be understood that DAPI is provided only as an example. Alternative dyes that enhance the visibility or contrast of the nucleus include, but are not limited to, Hoechst, propidium iodide, ethidium bromide, SYTOX Green, acridine orange, 7-aminoactinomycin D, T0-PR0-3 and SYTOX Red, and Y0PR0-1 and YOYO-1. The first stained image 504 (i.e. astrocyte stained with DAPI) is shown in Fig. 5. The second stained image enhances the visibility or contrast of the at least one cell's edges. In other words, the tissue sample is stained with an antibody that highlights the distribution and localization of a target protein, and the second stained image is acquired. Example antibodies include, but are not limited to, GFAP for astrocytes, NeuN for neurons, Olig2 for oligodendrocytes, and Desmin for muscle cells. The second stained image 506 (i.e. astrocyte stained with GFAP) is shown in Fig. 5.

[0064] At step 120, the method includes receiving a data matrix that represents a gene expression profile of the at least one cell. In some implementations, transcriptomic experiments are completed after cell staining experiments. For example, transcriptomic imaging can be completed after stained imaging. Optionally, in other implementations, transcriptomic and cell staining experiments can be completed simultaneously, for example using two different tissue sections of the same cell. In such implementations, the two different tissue sections are adjacent so the cellular compositions and morphologies are similar. It should be understood that the data matrix represents the gene expression profile of the same cell captured by the first and second stained images. An example data matrix 508 of the same cell captured by the first and second stained images 504, 506 is shown in Fig. 5. The data matrix can be obtained through single-cell spatial transcriptomics, which is a technique in molecular biology that allows researchers to study the gene expression profiles of individual cells within their spatial context in tissues. Single-cell spatial transcriptomics techniques aim to address this limitation by providing a high- resolution view of gene expression at the single-cell level while preserving the spatial information of cells within a tissue. This enables researchers to understand the spatial organization of diverse cell types and their interactions, contributing to a more comprehensive understanding of tissue function and development. Fig. 6 illustrates example techniques and platforms for obtaining the data matrix - Stereo-seq chip (Fig. 6, top), CosMx SMI spatial multiomics platform from NanoString Technologies, Inc. of Seattle, WA (Fig. 6, middle), andXenium single cell spatial imaging platform of lOx Genomics, Inc. of Pleasanton, CA (Fig. 6, bottom). Each technique and platform is discussed in the Examples below.

[0065] At step 130, the method includes combining the first stained image and the second stained image to create an overlay image. As used herein, image combination (sometimes referred to as “fusion”) involves merging information from multiple images into a single composite image. Optionally, in some implementations, the first and second stained images are registered (sometimes referred to as “alignment”) prior to combination. The overlay image contains the most relevant features from both the first stained image (e.g. cell nucleus) and the second stained image (e.g. cell edges). For example, with reference to Fig. 5, the first stained image 504 is combined with the second stained image 506 to obtain an overlay image 510. The goal is to enhance the information content or improve the visual representation by combining complementary data from different sources. Thus, image combination is used to highlight different aspects (e.g., the boundary of a cell or an organelle or the location of a protein) of the at least one cell, i.e. the nucleus visible in the first stained image 504 and the edges visible in the second stained image 506. The overlay image is used to identify the at least one cell within the tissue sample. For example, the visibility of a nucleus visible in the overlay image, which is present because of the first stained image, confirms the existence of the at least one cell in the second stained image.

[0066] At step 140, the method includes analyzing at least one of the second stained image or the overlay image to delineate a boundary of the at least one cell. In some implementations, as shown in Fig. 5, the second stained image is analyzed to delineate a boundary of the at least one cell. An example process is shown in Fig. 7. Optionally, the analysis can include: filtering the second stained image based on at least one of a signal intensity or a structure size; performing an edge detection process to identify a plurality of pixels that form the boundary of the at least one cell; retaining the plurality of pixels that form the boundary of the at least one cell, wherein the boundary of the at least one cell includes a plurality of gaps between pixels of the plurality of pixels; and connecting the boundary of the at least one cell by extending through the plurality of gaps. Optionally, the edge detection process is a modified Sobel edge detection technique, which can include any or all of the following modifications to a Sobel edge detection algorithm:

[0067] 1. Non-linear Edge score adjustment: After computing the edge score using theSobel filter, the edge score is non-linearly adjusted by raising it to the power of 0.8. This adjustment could enhance or suppress edge scores, affecting edge visibility.

[0068] 2. Conditional edge score scaling: If the edge score is >= 0.2, it can be further scaled by raising it to the power of 0.6. This conditional scaling can enhance lower scores, making the detection more sensitive to subtle edges.

[0069] 3. Two-stage Edge detection process: The algorithm applies a two-stage process for edge detection. Initially, it can compute and adjusts edge scores. In the second stage, it reapplies the Sobel operation on the output from the first stage without the non-linear adjustments. This iterative approach enhanced edge detection by refining the detected edges and highlighting certain features more prominently.

[0070] 4. Thresholding for final Edge detection: A hard threshold is applied to the second-stage output, setting pixels > 14 to 1 (indicating an edge) and all others to 0. This operation sharpens the edge map by converting it into a binary representation, thereby isolating prominent edges and removing less significant features and noise. As a result, the enhanced edge map distinctly delineates edges, offering a clearer and more defined structure suitable for further analysis.

[0071] Alternatively, in other implementations, the overlay image is analyzed to delineate a boundary of the at least one cell. For example, as shown in Fig. 11, an overlay image is used to determine the cell boundaries. The boundary of the cell in the bottom left corner of Fig. 11 was determined by the localization of the transcripts. The boundary of the cell in the bottom right corner of Fig. 11 was determined by the localization of the transcripts and histone antibody staining.

[0072] At step 150, the method includes creating a mask based on the boundary of the at least one cell in the at least one of the second stained image or the overlay image. As used herein, the mask identifies and delineates a region of interest, i.e. the boundary of the at least one cell. This disclosure contemplates using any known technique for creating the mask. One example technique is binary thresholding, where pixels in the second stained image are classified as foreground or background based on a specified intensity threshold. For example, pixels in the boundary of the at least one cell can be assigned one value (foreground) and the remaining pixels can be assigned another value (background).

[0073] At step 160, the method includes combining the mask and the data matrix to form a combined image for the at least one cell. As used herein, image combination (sometimes referred to as “fusion”) involves merging information from multiple images into a single composite image. Optionally, in some implementations, the mask and the data matrix are registered (sometimes referred to as “alignment”) prior to combination. An example combinedimage 512 for the at least one cell, e.g. the same cell captured by the first and second stained images 504, 506 and data matrix 508, is shown in Fig. 5. Thus, the combined image 512 captures both transcripts and staining. The combined image 512 contains the most relevant features from both the mask (e.g. the boundary of the cell) and the data matrix 508 (e.g. transcripts). As shown in Fig. 5, the boundary of the at least one cell (i.e. white lines) are visible in the combined image 512. Optionally, the method further includes generating display data for the combined image. As described below, as a result of more accurately identifying and delineating the boundary of the at least one cell, which has a complex shape, it is possible to more accurately capture the at least one cell’s transcriptome. This includes, but is not limited to, facilitating more accurate gene transcript counts for the at least one cell.

[0074] In some aspects, the method optionally further includes performing an omics analysis of the combined image obtained at step 160. For example, the omics analysis may be a transcriptomics analysis. Transcriptomics focuses on the study of the complete set of RNA transcripts (the transcriptome) produced by the genes in a particular cell or tissue. Optionally, the omics analysis includes determining transcript counts for the at least one cell. Alternatively or additionally, the omics analysis optionally includes determining a spatial distribution of the gene expression profile of the at least one cell. Optionally, the omics analysis is at least one of a cell typing analysis, a cell-cell communication analysis, an analysis of the number and distribution of different cell types, a differential gene expression analysis, a marker gene identification analysis, a protein-protein interaction network analysis, a cell trajectory analysis, or a functional enrichment analysis. It should be understood that other omics analyses than those described above can be performed. One example output of a transcriptomics analysis is a cell-by-gene matrix, for example as shown by Fig. 12.

[0075] Example results comparing the technique described with regard to Fig. 1 (referred to as “BrainSegmenter”) with a technique using a conventional cell segmentation method (e.g. Cellpose) are shown in Fig. 8. (A) Cellpose delineates the cell boundary by expanding from the nucleus' periphery outward, which results in a misrepresentation of astrocytes, depicting them erroneously as circular forms. (B) Cellpose fails to allocate transcripts to cells when the nucleus is on a different plane. (C) BrainSegmenter effectively outlines cells with stellate morphologies. (D) BrainSegmenter proficiently recognizes a cell and outlines its architecture, even when the nucleus is absent. (E) The process for categorizing cell types with BrainSegmenter incorporates a quality control (QC) phase to eliminate diminutive structures detected by BrainSegmenter. (F) Cellpose employs the nucleus morphology delineated by DAPIstaining as a basis for establishing cell boundaries, which overlooks the cell's elaborate branching structures, leading to incomplete transcript data capture. In contrast, BrainSegmenter maps the cellular architecture more accurately, retaining pertinent transcript information. BrainSegmenter and Segment Anything is abbreviated as BSegmenter and SMA, respectively. (G) Staining with GFAP and DAPI reveals that Cellpose struggles to identify cells when the nucleus is positioned on a distinct plane, resulting in an omission of data. Conversely, BrainSegmenter detects a cell and allocates transcripts accordingly. (H) Comparative heatmap of transcript counts obtained from cells delineated in (F) using Cellpose and BrainSegmenter. The gradient from blues to reds indicates the range from low to high transcript counts, respectively. This heatmap reveals a pronounced disparity in the number of transcripts detected for each gene, with BrainSegmenter producing elevated counts, indicating a more comprehensive capture of the cell's transcriptome. (I) Heatmap comparison of gene transcript counts obtained from the cell segmented in (G) using Cellpose and BrainSegmenter. The heatmap illustrates a marked difference in the quantity of transcripts detected for each gene, with Cellpose identifying no cell or genes.

[0076] In some aspects, the method optionally further includes performing a metabolomics analysis. In other words, the method optionally integrates metabolomics imaging with the transcriptomic analysis described above with regard to Fig. 1. In some implementations, metabolomics and transcriptomic experiments can be completed simultaneously, for example using two different tissue sections of the same cell. In such implementations, the two different tissue sections are adjacent so the cellular compositions and morphologies are similar. Optionally, in other implementations, transcriptomic experiments can be completed after metabolomics experiments. Optionally, in yet other implementations, transcriptomic experiments can be completed before metabolomics experiments. Metabolomics aims to profile and quantify the complete set of small molecules (metabolites) within a biological sample. It provides insights into the biochemical pathways and metabolic processes occurring in cells or organisms. Optionally, the method further includes receiving a mass spectrometry image that captures the at least one cell, wherein the mass spectrometry image is a spatial representation of a plurality of lipids or metabolites within the at least one cell. It should be understood that the mass spectrometry image is of the same cell captured by the first and second stained images as well as represented by the data matrix. A mass spectrometry image refers to a spatially resolved representation of the distribution of specific molecules or molecular ions across a sample.Imaging mass spectrometry is a powerful analytical technique that combines mass spectrometrywith spatial information, allowing researchers to visualize the spatial distribution of molecules in tissues, cells, or other sample types. One example mass spectrometry imaging technique is the matrix-assisted laser desorption / ionization time of flight (MALDI-TOF) technique. It should be understood that other IMS techniques may be used. Additionally, an example mass spectrometry image 902 is shown in Fig. 9. Optionally, the method further includes combining information associated with the combined image (e.g. the combined image 512 shown in Fig. 5) and the mass spectrometry image 902 to form an integrated image 904. The information associated with the combined image may be the boundary of the at least one cell and a type of the at least one cell. Optionally, in some implementations, the combined image and the mass spectrometry image are registered (sometimes referred to as “alignment”) prior to combination. The integrated image 904 contains the most relevant features from both the combined image (e.g. transcripts and cell boundary) and the mass spectrometry image (e.g. metabolites). Magnified image 904a and further magnified image 904b are also shown in Fig. 9. Segmented cells are visible in the further magnified image 904b. Optionally, the method further includes analyzing the integrated image to determine a metabolic profile of the at least one cell. Optionally, the method further includes determining a signal intensity and a location of at least one metabolite and a metabolic state of the at least one cell. Alternatively or additionally, the method further includes determining at least one of a cell type, a cell function, or a cell state of the at least one cell based on the metabolic profile.

[0077] hi some aspects, the techniques described herein relate to a method including: performing, for a biological sample from a subject, the method of single-cell spatial omics analysis as described above with regard to Fig. 1 ; and determining a diagnosis, prognosis, or treatment plan for the subject based on one or more respective cell-by-gene matrices for one or more cells of the biological sample. In some aspects, the method further includes administering a treatment to the subject based on the diagnosis, prognosis, or treatment plan. The method described with regard to Fig. 1 (referred to as “BrainSegmenter”) enables more accurate segmentation of cells with non-spherical shapes and can facilitate prognosis and the development of therapeutics. For example, BrainSegmenter can be used to analyze brain tumor tissue treated with laser ablation (or chemical treatment). If most brain tumor cells are undergoing apoptosis, it suggests that the therapy is effective. If the treated tumor tissue has a larger number of blood vessel cells (endothelial cells) than untreated tissue, a combined therapy with an angiogenesis inhibitor can be recommended and / or administered to inhibit the formation of blood vessels and treat the patients. Alternatively or additionally, BrainSegmenter can facilitate the development oftherapeutics involves seeding cells (possibly non-spherical) on a surface and treating them with a library of chemicals anchored or conjugated on another surface. Then, using BrainS egmenter and other omics analyses, the therapeutic effects of the chemicals on the cells (e.g., the toxicity of chemicals on cancer cells or the effect of the chemical in inducing stem cell differentiation) can be evaluated in a high-throughput manner.

[0078] In some aspects, the techniques described herein relate to a computer- implemented method of single-cell spatial omics analysis. In some implementations, as discussed above, the cell has a complex shape. In other implementations, the cell does not have a complex shape (e.g. round or elliptical cells). The computer-implemented method includes: receiving a plurality of images that capture at least one cell, the plurality of images including a first stained image and a second stained image, wherein the first stained image enhances the visibility or contrast of the at least one cell’s nucleus, and wherein the second stained image enhances the visibility or contrast of the at least one cell's edges; receiving a data matrix that represents a gene expression profile of the at least one cell; combining the first stained image and the second stained image to create an overlay image; analyzing at least one of the second stained image or the overlay image to delineate a boundary of the at least one cell; creating a mask based on the boundary of the at least one cell in the at least one of the second stained image or the overlay image; and combining the mask and the data matrix to form a combined image for the at least one cell. Additionally, the step of analyzing at least one of the second stained image or the overlay image to delineate the boundary of the at least one cell includes: filtering the at least one of the second stained image or the overlay image based on at least one of a signal intensity or a structure size; performing an edge detection process to identify a plurality of pixels that form the boundary of the at least one cell; retaining the plurality of pixels that form the boundary of the at least one cell, wherein the boundary of the at least one cell includes a plurality of gaps between pixels of the plurality of pixels; and connecting the boundary of the at least one cell by extending through the plurality of gaps.

[0079] Example Methods of Cell Segmentation

[0080] Referring now to Fig. 2, a flowchart illustrating an example method for cell segmentation is shown. This disclosure contemplates that the logical operations of Fig. 2 can be performed by a computing device. An example computing device is shown in Fig. 4. Segmenting cell boundaries is a critical task in biological image analysis, enabling researchers to quantify cell morphology, track cellular dynamics, and understand cellular structures in both healthy and disease states. This process involves distinguishing cells from the background and from eachother in an image, which can be challenging due to variability in cell shapes, sizes, and densities, as well as the presence of noise and overlapping structures in complex tissue environments.

[0081] At step 210, the method includes receiving a plurality of images that capture at least one cell having a complex shape. The images may capture a single cell or a plurality of cells. Additionally, the at least one cell may be part of a tissue sample. It should be understood that a tissue sample is provided only as an example. Alternatively, the at least one cell may be part of a cell culture, a cell suspension, or other group of cells. As noted above, the at least one cell has a complex shape. Existing automated cell segmentation techniques (e.g. Cellpose, which is an open-source anatomical segmentation algorithm) are unable to accurately segment cells having complex shapes as described in the Examples below. For example, such existing automated cell segmentation techniques tend to oversimplify cellular shapes, categorizing cells as either round or elliptical. This negatively impacts spatial cell omics analyses as described in the Examples below. Accordingly, as discussed above with regard to Fig. 1 , a complex shape is not a round or elliptical shape. Example complex shapes include, but are not limited to, elongated, cylindrical, polygonal spindle, pyramid, dome, or star shapes. Example cells having complex shapes include, but are not limited to, a neuronal cell (multipolar or bipolar with elongated axons and dendrites), an astrocyte (star-shaped), a fibroblast, a mesenchymal stem cell (spindle-like shape), a smooth muscle cell (spindle-shaped), a skeletal muscle fiber (cylindrical and multinucleated), an osteoclasts (dome-shaped and multinucleated), a keratinocyte (polygonal), an endothelial cell (elongated polygons, usually hexagonal), or a Sertoli cell (pyramid- shaped). In the Examples below, the at least one cell is a neuronal cell. It should be understood that a neuronal cell is provided only as an example cell having a complex shape.

[0082] The plurality of images comprise a first stained image and a second stained image. The first stained image and the second stained image capture the at least one cell having a complex shape. In other words, the at least one cell captured by the first and second stained images is the same cell. For example, the at least one cell may be part of a tissue sample that is fixed to a substrate. The tissue sample is then stained prior to imaging. Cell staining is a technique that involves the application of dyes or other markers to visualize specific cellular structures, specific proteins, or other biomolecules within cells or tissues. Staining is used to enhance the contrast and visibility of cells, allowing researchers to study their morphology, structure, and function.

[0083] An example cell 502 having a complex shape (an astrocyte) is shown in Fig. 5. It should be understood that an astrocyte is only provided as an example of a cell having acomplex shape. This disclosure contemplates using the method of Fig. 1 on other cells. The first stained image enhances the visibility or contrast of the at least one cell's nucleus. In other words, the tissue sample is stained with a dye that binds to genetic materials (e.g., DNA) found in the at least one cell’s nucleus, and the first stained image is acquired. DAPI is an example dye that enhances the visibility or contrast of the nucleus. It should be understood that DAPI is provided only as an example. Alternative dyes that enhance the visibility or contrast of the nucleus include, but are not limited to, Hoechst, propidium iodide, ethidium bromide, SYTOX Green, acridine orange, 7 -aminoactinomycin D, TO-PRO-3 and SYTOX Red, and Y0PR0-1 and YOYO-1. The first stained image 504 (i.e. astrocyte stained with DAPI) is shown in Fig. 5. The second stained image enhances the visibility or contrast of the at least one cell's edges. In other words, the tissue sample is stained with an antibody that highlights the distribution and localization of a target protein, and the second stained image is acquired. Example antibodies include, but are not limited to, GFAP for astrocytes, NeuN for neurons, Olig2 for oligodendrocytes, and Desmin for muscle cells. The second stained image 506 (i.e. astrocyte stained with GFAP) is shown in Fig. 5.

[0084] At step 220, the method includes combining the first stained image and the second stained image to create an overlay image. As discussed above with regard to Fig. 1, image combination (sometimes referred to as “fusion”) involves merging information from multiple images into a single composite image. Optionally, in some implementations, the first and second stained images are registered (sometimes referred to as “alignment”) prior to combination. The overlay image contains the most relevant features from both the first stained image (e.g. cell nucleus) and the second stained image (e.g. cell edges). For example, with reference to Fig. 5, the first stained image 504 is combined with the second stained image 506 to obtain an overlay image 510. The goal is to enhance the information content or improve the visual representation by combining complementary data from different sources. Thus, image combination is used to highlight different aspects of the at least one cell, i.e. the nucleus visible in the first stained image 504 and the edges visible in the second stained image 506. The overlay image is used to identify the at least one cell within the tissue sample. For example, the visibility of a nucleus visible in the overlay image, which is present because of the first stained image, confirms the existence of the at least one cell in the second stained image.

[0085] At step 230, the method includes analyzing at least one of the second stained image or the overlay image to delineate a boundary of the at least one cell. In some implementations, as shown in Fig. 5, the second stained image is analyzed to delineate aboundary of the at least one cell. This process is also described in the Examples. An example process is shown in Fig. 7. Optionally, the analysis can include: filtering the second stained image based on at least one of a signal intensity or a structure size; performing an edge detection process to identify a plurality of pixels that form the boundary of the at least one cell; retaining the plurality of pixels that form the boundary of the at least one cell, wherein the boundary of the at least one cell includes a plurality of gaps between pixels of the plurality of pixels; and connecting the boundary of the at least one cell by extending through the plurality of gaps. Optionally, the edge detection process is a modified Sobel edge detection technique described in Example 4, Appendix B. Alternatively, in other implementations, the overlay image is analyzed to delineate a boundary of the at least one cell. For example, as shown in Fig. 11, an overlay image is used to determine the cell boundaries. The boundary of the cell in the bottom left corner of Fig. 11 was determined by the localization of the transcripts. The boundary of the cell in the bottom right corner of Fig. 11 was determined by the localization of the transcripts and histone antibody staining.

[0086] At step 240, the method includes segmenting the at least one cell based on the boundary of the at least one cell in the at least one of the second stained image or the overlay image. One example segmentation technique is thresholding, where pixels in the second stained image are classified as foreground or background based on a specified intensity threshold. For example, pixels in the boundary of the at least one cell can be assigned one value (foreground) that is significantly different than the intensity values of the remaining pixels (background). Optionally, the method further includes generating display data for the segmented cell.

[0087] In some aspects, the techniques described herein relate to a computer- implemented method for cell segmentation. In some implementations, as discussed above, the cell has a complex shape. In other implementations, the cell does not have a complex shape (e.g. round or elliptical cells). The computer-implemented method includes: receiving a plurality of images that capture at least one cell, the plurality of images including a first stained image and a second stained image, wherein the first stained image enhances the visibility or contrast of the at least one cell’s nucleus, and wherein the second stained image enhances the visibility or contrast of the at least one cell's edges; combining the first stained image and the second stained image to create an overlay image; analyzing at least one of the second stained image or the overlay image to delineate a boundary of the at least one cell; and segmenting the at least one cell based on the boundary of the at least one cell in the at least one of the second stained image or the overlay image. Additionally, the step of analyzing at least one of the second stained image or the overlayimage to delineate the boundary of the at least one cell includes: filtering the at least one of the second stained image or the overlay image based on at least one of a signal intensity or a structure size; performing an edge detection process to identify a plurality of pixels that form the boundary of the at least one cell; retaining the plurality of pixels that form the boundary of the at least one cell, wherein the boundary of the at least one cell includes a plurality of gaps between pixels of the plurality of pixels; and connecting the boundary of the at least one cell by extending through the plurality of gaps.

[0088] Example Methods of Single-Cell Spatial Metabolomics

[0089] Referring now to Fig. 3, a flowchart illustrating an example method for singlecell spatial metabolomics analysis is shown. This disclosure contemplates that the logical operations of Fig. 3 can be performed by a computing device. An example computing device is shown in Fig. 4. Metabolomics aims to profile and quantify the complete set of small molecules (metabolites) within a biological sample. It provides insights into the biochemical pathways and metabolic processes occurring in cells or organisms.

[0090] At step 310, the method includes receiving a transcriptomic image for at least one cell, wherein the transcriptomic image is a spatial representation of a gene expression profile of the at least one cell. Optionally, in some implementations, the at least one cell has a complex shape as described herein, and the transcriptomic image is obtained according to the method described with respect to Fig. 1 , including segmentation for complex shapes and single-cell spatial transcriptomics. In other words, in such implementations, the transcriptomic image is a combined image (e.g. combined image 512 of Fig. 5). Alternatively, in other implementations, the transcriptomic image is obtained according to methods where cells are segmented using conventional techniques such as Cellpose and include single-cell spatial transcriptomics. In some implementations, transcriptomic experiments are completed after cell staining experiments. For example, transcriptomic imaging can be completed after stained imaging. Optionally, in other implementations, transcriptomic and cell staining experiments can be completed simultaneously, for example using two different tissue sections of the same cell. In such implementations, the two different tissue sections are adjacent so the cellular compositions and morphologies are similar. Fig. 6 illustrates example techniques and platforms for obtaining the data matrix - Stereo-seq chip (Fig. 6, top), CosMx SMI spatial multiomics platform from NanoString Technologies, Inc. of Seattle, WA (Fig. 6, middle), and Xenium single cell spatial imaging platform of lOx Genomics, Inc. of Pleasanton, CA (Fig. 6, bottom). Each technique and platform is discussed in the Examples below.

[0091] At step 320, the method includes receiving a mass spectrometry image that captures the at least one cell, wherein the mass spectrometry image is a spatial representation of a plurality of lipids or metabolites within the at least one cell. In some implementations, metabolomics and transcriptomic experiments can be completed simultaneously, for example using two different tissue sections of the same cell. In such implementations, the two different tissue sections are adjacent so the cellular compositions and morphologies are similar. Optionally, in other implementations, transcriptomic experiments can be completed after metabolomics experiments. Optionally, in yet other implementations, transcriptomic experiments can be completed before metabolomics experiments. A mass spectrometry image refers to a spatially resolved representation of the distribution of specific molecules or molecular ions across a sample. Imaging mass spectrometry is a powerful analytical technique that combines mass spectrometry with spatial information, allowing researchers to visualize the spatial distribution of molecules in tissues, cells, or other sample types. One example mass spectrometry imaging technique is the matrix- assisted laser desorption / ionization time of flight (MALDI-TOF) technique. It should be understood that other IMS techniques may be used. Additionally, an example mass spectrometry image 902 is shown in Fig. 9.

[0092] At step 330, the method includes combining information associated with the transcriptomic image and the mass spectrometry image 902 to form an integrated image 904. The information associated with the transcriptomic image may be the boundary of the at least one cell and a type of the at least one cell. Optionally, in some implementations, the transcriptomic image and the mass spectrometry image are registered (sometimes referred to as “alignment”) prior to combination. The integrated image 904 contains the most relevant features from both the transcriptomic image (e.g. transcripts, cell boundary, and cell type) and the mass spectrometry image (e.g. metabolites). Magnified image 904a and further magnified image 904b are also shown in Fig. 9. Segmented cells are visible in the further magnified image 904b.

[0093] At step 340, the method includes analyzing the integrated image to determine a metabolic profile of the at least one cell. For example, the method can include determining a signal intensity and a location of at least one metabolite and a metabolic state of the at least one cell. Alternatively or additionally, the method further includes determining at least one of a cell type, a cell function, or a cell state of the at least one cell based on the metabolic profile. Optionally, the method further includes generating display data for the integrated image. Fig. 10 illustrates an evaluation of the lipids in different cell types using the method described with regard to Fig. 3. (A-F) Overlay an image from imaging mass spectrometry (IMS) with a cellsegmentation image from single-cell spatial transcriptomics yields a single-cell spatial lipidomic image. The spatial distribution (A-B) and quantification (C-F) of SHexCer 42:2;O3 (m / z=906.5968) in oligodendrocytes (rhombus) and astrocytes (triangle). (G-L) The spatial distribution (G-H) and quantification (I-L) of PC 41 :5 (m / z=888.5872) in oligodendrocytes (star) and astrocytes (triangle). As, astrocyte; O, oligodendrocyte; En, endothelial cell; M, microglia; N, neuron; OPC, oligodendrocyte progenitor cells; P, pericyte; V, vascular leptomeningeal cells. Scale bars are 800 pm. Values are mean ± SD. p values determined by Mann- Whitney U tests. ***p < 0.001. *p < 0.05. Single-cell spatial metabolomics analysis is also illustrated by Example 5, Appendix C.

[0094] In some aspects, the method includes: performing, for a biological sample from a subject, the method for single-cell spatial metabolomics as described above with regard to Fig. 3; and determining a diagnosis, prognosis, or treatment plan for the subject based on one or more respective metabolic profiles for one or more cells of the biological sample. In some aspects, the method further includes administering a treatment to the subject based on the diagnosis, prognosis, or treatment plan.

[0095] Example Computing Device

[0096] It should be appreciated that the logical operations described herein with respect to the various figures may be implemented (1) as a sequence of computer implemented acts or program modules (i.e., software) running on a computing device (e.g., the computing device described in Fig. 4), (2) as interconnected machine logic circuits or circuit modules (i.e., hardware) within the computing device and / or (3) a combination of software and hardware of the computing device. Thus, the logical operations discussed herein are not limited to any specific combination of hardware and software. The implementation is a matter of choice dependent on the performance and other requirements of the computing device. Accordingly, the logical operations described herein are referred to variously as operations, structural devices, acts, or modules. These operations, structural devices, acts and modules may be implemented in software, in firmware, in special purpose digital logic, and any combination thereof. It should also be appreciated that more or fewer operations may be performed than shown in the figures and described herein. These operations may also be performed in a different order than those described herein.

[0097] Referring to Fig. 4, an example computing device 400 upon which the methods described herein may be implemented is illustrated. It should be understood that the example computing device 400 is only one example of a suitable computing environment upon which themethods described herein may be implemented. Optionally, the computing device 400 can be a well-known computing system including, but not limited to, personal computers, servers, handheld or laptop devices, multiprocessor systems, microprocessor-based systems, network personal computers (PCs), minicomputers, mainframe computers, embedded systems, and / or distributed computing environments including a plurality of any of the above systems or devices. Distributed computing environments enable remote computing devices, which are connected to a communication network or other data transmission medium, to perform various tasks. In the distributed computing environment, the program modules, applications, and other data may be stored on local and / or remote computer storage media.

[0098] In its most basic configuration, computing device 400 typically includes at least one processing unit 406 and system memory 404. Depending on the exact configuration and type of computing device, system memory 404 may be volatile (such as random access memory (RAM)), non-volatile (such as read-only memory (ROM), flash memory, etc.), or some combination of the two. This most basic configuration is illustrated in Fig. 4 by box 402. The processing unit 406 may be a standard programmable processor that performs arithmetic and logic operations necessary for operation of the computing device 400. The computing device 400 may also include a bus or other communication mechanism for communicating information among various components of the computing device 400.

[0099] Computing device 400 may have additional features / functionality. For example, computing device 400 may include additional storage such as removable storage 408 and non-removable storage 410 including, but not limited to, magnetic or optical disks or tapes. Computing device 400 may also contain network connection(s) 416 that allow the device to communicate with other devices. Computing device 400 may also have input device(s) 414 such as a keyboard, mouse, touch screen, etc. Output device(s) 412 such as a display, speakers, printer, etc. may also be included. The additional devices may be connected to the bus in order to facilitate communication of data among the components of the computing device 400. All these devices are well known in the art and need not be discussed at length here.

[0100] The processing unit 406 may be configured to execute program code encoded in tangible, computer-readable media. Tangible, computer-readable media refers to any media that is capable of providing data that causes the computing device 400 (i.e., a machine) to operate in a particular fashion. Various computer-readable media may be utilized to provide instructions to the processing unit 406 for execution. Example tangible, computer-readable media may include, but is not limited to, volatile media, non-volatile media, removable mediaand non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. System memory 404, removable storage 408, and non-removable storage 410 are all examples of tangible, computer storage media. Example tangible, computer-readable recording media include, but are not limited to, an integrated circuit (e.g., field-programmable gate array or application-specific IC), a hard disk, an optical disk, a magneto-optical disk, a floppy disk, a magnetic tape, a holographic storage medium, a solid-state device, RAM, ROM, electrically erasable program read-only memory (EEPROM), flash memory or other memory technology, CD-ROM, digital versatile disks (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices.

[0101] In an example implementation, the processing unit 406 may execute program code stored in the system memory 404. For example, the bus may carry data to the system memory 404, from which the processing unit 406 receives and executes instructions. The data received by the system memory 404 may optionally be stored on the removable storage 408 or the non-removable storage 410 before or after execution by the processing unit 406.

[0102] It should be understood that the various techniques described herein may be implemented in connection with hardware or software or, where appropriate, with a combination thereof. Thus, the methods and apparatuses of the presently disclosed subject matter, or certain aspects or portions thereof, may take the form of program code (i.e., instructions) embodied in tangible media, such as floppy diskettes, CD-ROMs, hard drives, or any other machine-readable storage medium wherein, when the program code is loaded into and executed by a machine, such as a computing device, the machine becomes an apparatus for practicing the presently disclosed subject matter. In the case of program code execution on programmable computers, the computing device generally includes a processor, a storage medium readable by the processor (including volatile and non-volatile memory and / or storage elements), at least one input device, and at least one output device. One or more programs may implement or utilize the processes described in connection with the presently disclosed subject matter, e.g., through the use of an application programming interface (API), reusable controls, or the like. Such programs may be implemented in a high level procedural or object-oriented programming language to communicate with a computer system. However, the program(s) can be implemented in assembly or machine language, if desired. In any case, the language may be a compiled or interpreted language and it may be combined with hardware implementations.

[0103] Examples

[0104] The following examples are put forth so as to provide those of ordinary skill in the art with a complete disclosure and description of how the compounds, compositions, articles, devices and / or methods claimed herein are made and evaluated, and are intended to be purely exemplary and are not intended to limit the disclosure. Efforts have been made to ensure accuracy with respect to numbers (e.g., amounts, temperature, etc.), but some errors and deviations should be accounted for. Unless indicated otherwise, parts are parts by weight, temperature is in °C or is at ambient temperature, and pressure is at or near atmospheric.

[0105] Example 1

[0106] A cell segmentation method, BrainSegmenter, for the accurate analysis of non-spherical cells within the central nervous system has been developed. Unlike existing technologies such as Stereoseq by STOmics, CosMx by NanoString, Xenium by lOx Genomics, and MERSCOPE by Vizgen, which primarily rely on Cellpose-based methods trained predominantly on non-neuronal cells, BrainSegmenter excels in identifying and analyzing the complex shapes of neural cells.

[0107] BrainSegmenter is tailored for the unique morphologies of brain cells, acknowledging their non-circular geometry. It utilizes specific antibody staining markers like GFAP, NeuN, and Olig2 to identify various cell types with high fidelity.

[0108] Current market solutions tend to oversimplify cell shapes into circular forms, which are not representative of cells in the brain. BrainSegmenter addresses this limitation by providing a segmentation framework that recognizes and adapts to the star-shaped astrocytes, as well as the multipolar nature of neurons and glial cells.

[0109] Advantages of BrainSegmenter include:

[0110] Enhanced accuracy in cell segmentation leads to more precise single-cell spatial transcriptomic analyses.

[0111] Facilitates in-depth studies of neurological development and disorders, where cell shape and type play a critical role in understanding disease mechanisms.

[0112] Offers a significant improvement over existing methods, which could lead to better-targeted therapeutics and diagnostics in neurology.

[0113] BrainSegmenter relates to a method and system for the segmentation of cells for the purpose of single-cell spatial omics analyses. BrainSegmenter enables the identification and analysis of individual cells within a heterogeneous sample, allowing for detailed examination of cellular composition and function.

[0114] Step 1: Image Overlay Creation

[0115] The process begins with the generation of overlaid images. This is achieved by staining DNA with fluorescent dyes such as DAPI or Hoechst, and RNA with markers for ribosomal RNA. Concurrently, antibody staining is employed to define cell boundaries: GFAP for astrocytes, NeuN for neurons, and Olig2 for oligodendrocytes.

[0116] Step 2: Cell Boundary Definition

[0117] Using the images obtained from Step 1, the invention utilizes the OpenCV software package along with algorithms described herein to delineate cell boundaries accurately. This step distinguishes individual cells in the subsequent analysis.

[0118] Step 3: Mask File Creation

[0119] Based on the defined cell boundaries, a mask file is created. This file is a digital representation that isolates each cell within the image, allowing for specific targeting in the analysis.

[0120] Step 4: Integration with Expression Matrix

[0121] The mask file is then integrated with a pre-existing expression matrix. This matrix contains data on gene expression levels within the sample. By aligning the mask file with the expression matrix, transcripts can be assigned to individual cells with high precision.

[0122] Resultant File for Downstream Analysis

[0123] The final output is a file that associates transcripts with specific cells. This file is an essential tool for downstream single-cell spatial omics analyses, which may include studying cell behavior, interaction, and function in a spatial context.

[0124] Example 2

[0125] Described below is a segmentation approach, MetaSegmentation, tailored for single-cell spatial metabolomics, a field hitherto only explored in its constituent parts. While the domain of single-cell metabolomics has been previously investigated, and high-resolution metabolomics has been studied, MetaSegmentation marks a pioneering stride in integrating these disciplines. Specifically, MetaSegmentation enables the precise overlay of a cell segmentation map, derived from single-cell transcriptomics data, with a complementary image from imaging mass spectrometry (IMS) data.

[0126] The transcriptomic data delineates the cellular and nuclear boundaries, transcript quantities, and locations with remarkable accuracy, whereas the IMS data provides exhaustive details regarding the quantity and spatial distribution of metabolites within the cells. The fusion of these two sophisticated imaging techniques permits a previously unattainableexamination of metabolites across varied cell types in both pathological and healthy tissue samples. This dualistic approach not only enhances understanding of the cellular landscape within a given tissue but also reveals metabolic alterations and mechanisms that may contribute to the pathogenesis of diseases.

[0127] MetaSegmentation delivers insights into the cellular and molecular orchestra that defines the state of health and the onset of disease, thereby offering a potent tool for biomedical research and therapeutic development.

[0128] Initial Spatial Transcriptomic Analysis:

[0129] The process commences with a single-cell spatial transcriptome assessment. Here, individual cell perimeters are demarcated utilizing nuclear staining, which involves extending the boundary of the nuclei by a range of 1 -20 microns outwardly. Alternatively, segmentation techniques may be employed based on data derived from antibody staining. Cell classification is achieved through the analysis of gene expression levels. The spatial transcriptomics techniques applicable include, but are not limited to, Stereoseq by STOmics, CosMx by NanoString, Xenium by lOx Genomics, and MERSCOPE by Vizgen.

[0130] Imaging Mass Spectrometry (IMS) Application:

[0131] Subsequent to the transcriptomic analysis, the method proceeds with imaging mass spectrometry on either an adjacent tissue section or the same section utilized in the preceding step. This technique elucidates the spatial distribution and concentration of various lipid entities or metabolites, which may encompass peptides and other chemical constituents, at a subcellular resolution. Quantitative and locational data acquisition is facilitated through analytical software platforms such as SCiLS Lab.

[0132] Data Integration and Visualization:

[0133] The culmination of the method involves the integration of the spatial transcriptomics and IMS data. This integrative approach enables the precise delineation of cell boundaries and the identification of cell types associated with distinct metabolites.

[0134] Example 3 TransMetaSegmentation (TMS) enables spatially resolved single-cell metabolomics:

[0135] Inherited metabolomic diseases (IMDs) encompass over 700 rare conditions that, although individually uncommon, collectively impact millions of individuals globally, imposing a significant burden on healthcare systems. For example, Gaucher disease, a lysosomal storage disorder, occurs in about 1.5 out of every 100,000 individuals, with enzyme replacement therapy costing over €5.7 million per patient over a lifetime. Another example isphenylketonuria (PKU), a disorder affecting phenylalanine metabolism, which occurs in approximately 1 in 10,000 individuals in Europe, with annual treatment expenses ranging from $19,000 to over $50,000 per patient. Nearly all diseases are associated with metabolomic disruptions. For instance, diabetes, which affected an estimated 463 million people worldwide in 2019, involves altered amino acid, fatty acid, and glucose metabolism, contributing to a global healthcare cost of USD 727 billion in 2017. The pervasive nature of metabolic alterations across diverse pathologies highlights the indispensable role of metabolomics in elucidating disease mechanisms.

[0136] To advance the understanding of disease mechanisms and discover biomarkers for diagnosis and prognosis, technologies that evaluate metabolites are crucial. One such technology is matrix- as sis ted laser desorption / ionization mass spectrometry imaging (MALDI MSI). MALDI MSI is a powerful analytical tool that enables the visualization and comparison of relative abundances of metabolites across samples, shedding light on biological processes and disease mechanisms. MALDI MSI uncovers intricate distribution patterns of a vast array of molecular species, including metabolites, lipids, peptides, proteins, and glycans. It can detect metabolites across a broad molecular weight spectrum, from 0 to 100,000 Daltons, with spatial resolutions ranging from approximately 1 to 100 pm. This versatility has made MALDI MSI invaluable for extensive metabolic profiling in complex tissues, offering insights into the interplay of biomolecules and their roles in physiological processes, disease mechanisms, and therapeutic interventions.

[0137] Techniques such as scSpatMet enable the determination of cell boundaries and cell types through staining with 35 cell marker antibodies. Yet, distinguishing subpopulations of cells in the brain, such as subclusters of neurons, astrocytes, and oligodendrocytes, remains challenging using antibodies. MALDI MSI analysis has a significant limitation: the difficulty in defining cell boundaries hinders its ability to determine cell typespecific metabolic profiles and unveil cellular functions and molecular interactions within complex tissues. To tackle this issue, single-cell spatial transcriptomics can be leveraged. These technologies provide detailed information on cell boundaries and cell types, overcoming the constraints of MALDI MSI. Although fluorescent in situ hybridization (FISH) was invented in the 1970s and spatially resolved transcriptomics has been around for many years, technologies capable of true single-cell resolution only became commercially available in 2022. Following the commercial launch of these technologies, rapid advancements have been made, particularly in the number of genes that can be evaluated and the variety of sample types. For instance, NanoString'sCosMx platform, which previously could assess approximately 960 genes, now features a 6,000- gene panel, with plans to expand to an 18,000-gene panel by 2025, enabling detailed transcriptomic analysis of human and mouse samples at a resolution of 200 nm. Similarly, 10X Genomics' Xenium platform, which initially could evaluate 248 genes, now offers 5,000-gene panels with the same 200 nm resolution for both human and mouse tissues. Additionally, STOmics has introduced the Stereo-seq OMNI, allowing for comprehensive transcriptomic evaluation at a resolution of 500 nm or 710 nm in formalin-fixed paraffin-embedded (FFPE) tissues, whereas the technology released in 2022, Stereo-seq, could only evaluate fresh frozen tissue. All three spatial transcriptomics technologies offer promising solutions to address the challenges of cell boundary definition and cell typing in MALDI MSI.

[0138] Most single-cell spatial transcriptomics technologies rely on Cellpose for cell segmentation, which uses deep learning and vector flow representation to define cell boundaries. This algorithm converts cell masks into vector flow fields, which guide pixels toward cell centers through simulated diffusion-based spatial gradients. At the core of Cellpose is a U- Net-based neural network trained to predict these gradients. The network outputs horizontal and vertical gradient maps, along with a binary inside / outside map. During the inference phase, pixels "flow" along the predicted gradients, converging at fixed points that represent cell centers. This process effectively groups pixels into regions of interest (ROIs), delineating individual cells. The binary map plays a crucial role in refining the masks by excluding pixels outside cell boundaries. Cellpose's approach enables the capture of diverse cell morphologies without relying on predetermined assumptions about cell shapes or sizes.

[0139] Following cell segmentation, various approaches can be applied for cell typing. One approach for cell typing is through unsupervised machine learning techniques. Methods such as clustering algorithms (e.g., k-means, hierarchical clustering) and community detection methods (e.g., Louvain or Leiden) are applied to high-dimensional gene expression data to group cells with similar transcriptomic profiles. These methods are often coupled with dimensionality reduction techniques like PCA, t-SNE, or UMAP to visualize and interpret the resulting cell clusters. Unsupervised approaches can be valuable for discovering novel cell types or states without prior assumptions, as they can reveal hidden patterns in complex tissues or developmental processes. Additionally, manual annotation based on cell type marker genes remains a crucial step in cell typing analysis, serving as a validation tool and providing expert biological insight. To complement unsupervised methods, supervised machine learning algorithms like InSituType have been developed to leverage existing knowledge of cell types formore precise classification. These methods use labeled training data to build models that can accurately identify known cell types based on their gene expression signatures and spatial context. By integrating spatial information with transcriptomic data, supervised approaches can improve the accuracy of cell type identification, especially in complex tissues with well-defined cellular populations. Researchers often use a combination of these approaches, integrating machine learning-derived clusters with manual curation to achieve comprehensive and accurate cell type annotations in spatial transcriptomics studies. This study used Xenium to provide the cell boundary and cell type information for MALDI MSI and achieve single-cell spatial metabolomics

[0140] Result

[0141] Single-cell spatial MSI is achieved by combining single-cell spatial transcriptomic and high-resolution MSI. To analyze metabolites and transcripts from the same cells, the study employed an open-book mounting technique (Figure 13A). Using Xenium technology, which involves rolling circle amplification followed by 15 cycles of hybridization, imaging, and washing, hundreds to thousands of genes can be evaluated with subcellular resolution (Figure 13B). Simultaneously, MALDI MSI allows for the spatial profiling of hundreds of metabolites at a square pixel size of 100 pm^ (Figure 13C). By integrating Xenium with MALDI MSI, the study achieved single-cell spatial metabolomics (Figure 13D), a technique referred to as TransMetaSegmentation (TMS), an alternative segmentation and cell typing method that integrates MALDI MSI molecular data with single-cell spatial transcriptomic analysis. This approach not only delineates cell boundaries and defines cell types based on a number of marker genes but also effectively allocates metabolites to specific cell types in a high- throughput manner. TMS involves preparing two serial brain sections: one for MALDI MSI and the other for Xenium analysis. Integrating data from both modalities enables high-throughput assignment of metabolites to specific cell types (Figure 13). TMS uncovers the spatial relationships between transcripts and metabolites at single-cell resolution, providing a powerful tool for unraveling complex biological processes within tissue

[0142] Fiducial markers and co-registration

[0143] In some implementations of the present disclosure, incisions and fiducial markers can be used to improve co-registration accuracy. In an example implementation, four incisions or fiducial markers are placed at the edges of the tissue. The incisions are minimized to less than 5 pm whenever possible, as larger incisions — such as 100 pm — can introduce errors.Given that a single cell measures only 10-20 pm, misalignment can compromise spatial accuracy.

[0144] Two consecutive sections are prepared before mounting the first section. The MSI section can be mounted first, followed by the Stereo-seq section. If the MSI section is mounted incorrectly, two new sections can still be prepared. However, if the Stereo-seq section is mounted first and an error occurs with either section, it results in the loss of an expensive Stereo-seq chip, as once the Stereo-seq section is placed, it cannot be repositioned or replaced.

[0145] When mounting the tissue onto slides or chips used for MSI or spatial transcriptomics, the slide or chip must not be pressed onto the tissue. Instead, forceps should be used to lift and transfer the tissue onto the chip. Pressing the chip onto the tissue can cause RNA to adhere to the metal surface of the cryotome, leading to uneven transcriptomic signal distribution, as observed in previous Stereo-seq images. Additionally, improper contact may disrupt the uniform distribution of nanoballs on the chip.

[0146] A transgenic mouse model recapitulates the Canavan disease. This study utilized brain tissue from the Aspanur^ mouse model, which recapitulates the features of Canavan disease. This transgenic strain harbors an ENU-induced Q193X nonsense mutation in the aspartoacylase (Aspa) gene, resulting in a loss-of-function allele that closely mimics the human disorder. Consistent with the clinical presentation of Canavan disease, these mice exhibited early-onset spongy degeneration of CNS myelin, elevated N-acetyl aspartate (NAA) levels, and progressive neurological symptoms, including ataxia, tremors, and seizures. Histological and biochemical analyses revealed reduced myelin protein expression, astrogliosis, and cortical neuron loss. Using TMS, dysregulated lipid profiles were observed across various cell types in Canavan mouse brains. The successful application of TMS in this context not only deepens the understanding of Canavan disease pathophysiology but also serves as a proof of concept for its broader utility in studying a wide range of neurological disorders characterized by metabolic dysfunction.

[0147] Cellular segmentation is conducted using the Xenium platform. The Xenium platform employs a multi-step approach for cell segmentation. The process begins with nucleus detection using DAPI staining and a custom neural network trained on diverse tissue types. For samples prepared with Cell Segmentation Staining, Xenium utilizes a multimodal cell segmentation algorithm that incorporates deep learning models to analyze multi-channel stain images, including cell boundary, cell interior, and nuclear stains. As a result, the cell boundaries in the mouse brain tissue can be delineated (Figures 14A-14C). In cases where staining isunavailable or insufficient, the system defaults to a nuclear expansion method, extending boundaries by 5-15 pm or until encountering another cell. The segmentation process is further refined by integrating additional data, such as fluorescence intensity and spatial location. Automated quality control checks are implemented to identify and correct segmentation errors, such as misidentified cell boundaries or incomplete cell segmentation. Additionally, the method incorporates manual review steps, allowing trained experts to inspect and adjust segmented images.

[0148] Xenium determines cell type based on gene expression patterns.Xenium analysis of the hippocampal region revealed a rich cellular landscape, detecting 91,023,571 transcripts across 285,830 cells. Unsupervised clustering identified 53 distinct cell populations (Figure 14D), with clusters 1 and 2 representing 7% and 5% of the total cells, respectively (Figure 14G). The spatial distribution of these clusters is illustrated in Figure 14E. Cell type annotation, based on transcriptomic profiles, was performed using established marker genes (Figure 14F). This approach facilitated the identification and characterization of multiple cell types, such as astrocytes, endothelial cells, perivascular macrophages, and vascular leptomeningeal cells, among others. Oligodendrocytes were identified using Gfap, Gjc3, Opalin, SoxlO, and Zip536; neurons by Clmn, Nelli, Proxl, Rims3, and Slcl7a6; endothelial cells by Aldhla2, Coll al, Den, Igf2, and Pdgfra; and astrocytes by Acsbgl, Aqp4, Id2, Ntsr2, and Slc39al2 (Figure 14H-J). Additionally, Xenium uncovered differential gene expression patterns between Canavan disease and control brains (Figure 18), elucidating the molecular underpinnings of this disorder.

[0149] Dysregulated metabolites in murine model brains unveiled by MALDI MSI. Using MALDI MSI, more than 120 lipids were putatively identified in mouse brains, with mass ranging from m / z value 50 to 2000, at a square pixel size of 100 pm^. These lipids span several categories, including phospholipids (e.g., phosphatidylethanolamines, PE), phosphatidic acids, and sphingolipids, e.g., hexosylceramides (HexCer), sulfated hexosylceramides (SHexCer), and ceramide phosphoethanolamines (CerPE) (Figure 15). MALDI MSI data unveils elevated levels of certain lipids in a Canavan mouse brain such as [HexCer(42:2;O3)-H]‘ (Figure 15F) and [SHexCer(36:l ;O3)-H]‘ (Figure 151), alongside reduced levels of others like [PE(O- 36:3)-H]’ (Figure 15A), [HexCer(42:2;O2)-H]- (Figure 15G) and [SHexCer(36:l;O2)-H]- (Figure 15H).

[0150] Co-registration of MALDI MSI and spatial transcriptomic images.Linear transformation and non-linear transformation have been employed to align multi-modal images depending on the type of deformation present. An affine transformation is effective for addressing global distortions while preserving important geometric properties. For more complex cases involving local or global distortion, non-linear transformations like B-spline transformations can be employed. This study also provided workflows involving affine transformation (Figure 16) and B-spline transformation (see Methods). The study employed Canny edge detection for the autonomous delineation of cell boundaries (Figures 16B-16C). When quantifying metabolite signals in individual cells, some methods exclude a pixel if less than half of its area falls within the selected ROI and assign full intensity to the ROI if more than half of the pixel area is included. Various methods have been developed for quantifying signal intensity in individual cells to ensure accuracy. This method calculates the total signal for each cell by determining the average signal intensity within the cell and multiplying it by the cell boundary area (See Methods)

[0151] TMS combines cellular information from Xenium with metabolite information from MALDI MSI. While MALDI MSI sheds light on the upregulation and downregulation of lipids, it lacks the ability to identify specific cell types associated with these metabolic changes. To overcome this limitation, the study leveraged Xenium and developed the TMS (Figure 19). The TMS workflow began with the visualization of the single-cell spatial transcriptomic image in Xenium Explorer 3, which provided a high-resolution map of gene expression across the tissue section. This image was saved and imported into SCiLS Lab, where it was aligned with the MALDI MSI image. To ensure accurate alignment, two anchor points were set on both the transcriptomic and MALDI MSI images. Following alignment, the transparency of the overlaid images was adjusted to optimize the visualization of the integrated image and examination of both transcriptomic and metabolomic features within the same spatial context. To facilitate cell-specific analysis, cell boundaries were delineated on the aligned images, creating regions of interest (ROIs) that corresponded to individual cells identified by Xenium. Subsequently, the metabolite signals within each ROI are quantified and analyzed statistically.

[0152] Attributing the dysregulated metabolites to specific cell types. Using TMS, the lipid levels were analyzed in identical cell types in control and Canavan brains, focusing on oligodendrocytes and astrocytes. These findings reveal significant alterations in lipid composition associated with Canavan disease. Specifically, both oligodendrocytes and astrocytes in Canavan’s brains exhibited elevated levels of [SHexCer(42: l;O3)-H]‘ (Figure 16A-16F) and[PI(40:6)-H] (Figure 17G-17L), alongside a marked reduction in [SHexCer(42:2;O2)-H] (Figure 16G-16L). Furthermore, astrocytes alone displayed increased levels of [PE(38:5)-H]’ (Figure 17E). Although these lipids have not been reported in the context of Canavan disease, they belong to broader lipid categories known to be dysregulated in various neurological disorders. Sphingolipids, including sulfated hexosylceramides (SHexCer), are critical components of the myelin sheath, where they maintain myelin integrity and facilitate axonmyelin interactions. Dysregulation of SHexCer levels has been associated with neurodegenerative and demyelinating disorders such as multiple sclerosis (MS) and Alzheimer’s disease (AD). In MS, elevated SHexCer levels exacerbate inflammation by activating microglia and astrocytes, leading to further myelin damage. Conversely, decreased SHexCer levels have been observed in AD using MALDI-MSI, which are associated with myelin damage. In Canavan disease, the elevated levels of [SHexCer(42: 1;O3)-H]‘ and reduced levels of[SHexCer(42:2;O2)-H]" point to similar mechanisms, potentially affecting oligodendrocyte and astrocyte function and leading to impaired myelination.

[0153] Glycerophospholipids (GPs) are another class of lipids implicated in neurological disorders. These lipids are essential for maintaining cellular membrane integrity, facilitating intracellular signaling, and supporting brain metabolism. Disruptions in glycerophospholipid metabolism have been linked to the pathogenesis of neurodegenerative diseases, including MS and AD. For example, alterations in phosphatidylcholines (PC) and phosphatidylethanolamines (PE) have been documented in MS, and experimental autoimmune encephalomyelitis (EAE), a model of MS, reveals significant changes in phosphatidylinositol (PI) levels. In this study, increased levels of [PI(40:6)-H]‘ and [PE(38:5)-H]‘ in astrocytes of Canavan disease brains suggest dysregulation in these lipid classes. Elevated PI levels may reflect altered membrane dynamics and intracellular signaling, while increased PE levels might indicate compensatory responses to maintain membrane integrity under disease-induced stress. [PI(40:6)-H]’ is a glycerophosphoinositol (PI), a subclass of glycerophospholipids. These lipids are crucial for cellular functions, including synaptic signaling and membrane maintenance, and disruptions in PI metabolism have been implicated in neuroinflammatory and neurodegenerative processes. The changes in SHexCer levels in Canavan disease echo findings in other disorders where sulfatide dysregulation leads to impaired myelination and neuronal degeneration. These findings highlight the importance of lipid dysregulation in the pathophysiology ofCanavan disease and showcase the capability of TMS in elucidating cell-type-specific metabolic changes in neurodegenerative conditions.

[0154] Discussion. Using MALDI MSI alone, the study revealed prominent dysregulation in lipid composition, including elevated levels of certain SHexCer, PG, and PS species and reduced levels of various PC species in Canavan brains. These findings underscore the profound metabolic disruptions associated with Canavan disease. However, MALDI MSI lacks cell boundary and cell type information. To address this problem, single-cell spatial transcriptomics is combined with MALDI MSI. This approach allows us to map the distribution of specific lipids within individual cells, unveiling elevated levels of some SHexCer species in oligodendrocytes and astrocytes and an increased abundance of PG lipid species in oligodendrocytes and neurons in Canavan brains. Although this study focused solely on the spatial distribution of lipidomics, MALDLMSI and TMS can also be utilized to evaluate other metabolites, such as glycans and peptides, which are critical for understanding disease mechanisms. For example, one study correlated specific glycan and lipid signatures with distinct kidney cancer regions, helping to identify potential biomarkers for renal cell carcinoma and providing a deeper understanding of disease progression. In another study, the distribution patterns of N-glycans, peptides, and lipids were associated with specific pathological features in the lung, offering essential insights for elucidating disease mechanisms and identifying novel diagnostic markers or therapeutic targets in lung pathologies. Changes in N-glycan profiles were also observed in irradiated lung tissue, where correlations between glycan profiles and specific pathological changes were noted. This approach has uncovered potential markers for various radiation-induced tissue states, including fibrosis, edema, and macrophage infiltration, potentially opening new avenues for therapeutic interventions.

[0155] There are limitations to using TMS for defining cell boundaries and determining cell types. One limitation is the accuracy of cell segmentation, which relies on Cellpose 2.0. A notable drawback of Cellpose 2.0 is that it is less accurate when segmenting starshaped astrocytes and bipolar neurons. Including a more diverse training dataset from a broader range of cell types and employing deep learning methods to analyze cell transcripts could significantly improve the segmentation accuracy of Cellpose.

[0156] Some types of transcriptomics can have limits for cell typing. First, RNA levels may not always correlate perfectly with protein levels, and / or some antibody-based methods may be more accurate in defining cell boundaries. DNA staining, combined with algorithms such as Cellpose, can be employed to define cell boundaries in spatial transcriptomicstechnologies. These methods typically identify the nuclear boundary and extend it outward by 5- 15 pm to approximate the cell boundary. While this approach is generally effective for round or elliptical cells, it poses significant challenges when segmenting complex cellular morphologies in the brain, such as neurons with elongated axons or astrocytes with their characteristic starshaped structures. However, current antibody-based methods are limited in detecting hundreds of proteins. Single-cell spatial transcriptomic technologies (e.g., Stereo-seq, CosMx, Xenium) can capture thousands of transcripts. This allows for the identification of novel cell types, subtypes, transient cell states, and developmental trajectories that might be missed by current single-cell spatial proteomics analyses. For example, currently, Xenium can evaluate 5,000 genes in one experiment, CosMx can assess 6,000 genes in one experiment, and Bruker plans to release an 18,000-gene panel in 2025. Stereo-seq can measure 37,000 genes in one experiment. Currently, single-cell spatial proteomic technologies can determine cell shape better but are not yet on par with transcriptomics in terms of scalability. For example, MACSima Imaging Cyclic Staining (MICS) can apply 300 antibodies to the same specimen. CODEX can employ 104 antibodies, and CosMx can process approximately 64 antibodies in one experiment. Cyclic immunofluorescence (CycIF) techniques, such as t-CyCIF, enable imaging of up to 60 protein targets on a single tissue sample, while CyTOF, or mass cytometry, uses heavy metal isotope-labeled antibodies to profile over 40 parameters per cell. A combination of laser cutting and timsTOF Ultra 2 (detection limit is 31 pg of protein) can be used for spatial proteomics. However, a HEK293 cell contains around 19 pg of proteins, so laser cutting can be used to obtain proteins from the same cell type. This can be a promising approach to achieving single-cell spatial proteomics, and can be challenging to process large numbers of cells efficiently.

[0157] While this study focused on using Canavan disease and Xenium (Xenium Explorer) as illustrative examples, other single-cell spatial technologies such as CosMx (napari) and MERSCOPE (MERSCOPE Vizualizer) can also be combined with MALDI MSI data to investigate the metabolic changes in various pathological conditions. In summary, TMS represents a powerful approach that enhances the functionality of MALDI MSI, enabling metabolomic analyses at a single-cell resolution. It advances the understanding of the metabolic changes and molecular interactions within different cell types across various diseases, contributing to the development of more effective diagnostic and prognostic tools.

[0158] Materials and Method Chemicals

[0159] Animal experiments. All animal procedures were approved by TheUniversity of California, Davis, Institutional Animal Care and Use Committee (IACUC). Canavan mice (Strain 008607) were purchased from Jackson Laboratory.

[0160] MALDI MSI. Mouse brains were dissected and then placed on a foil paper boat, which was then filled with pulverized dry ice. Isopentane (VWR, Cat. JTQ223-8) was poured over the dry ice, ensuring rapid temperature reduction and efficient freezing of the tissue to preserve its molecular and structural integrity. Once frozen, the brain was embedded in 2.6% carboxymethyl cellulose (CMC; EMD Millipore Corp., Burlington, MA), sectioned into 10 pm thick sections using a cryomicrotome (Leica Biosystems, Wetzlar, Germany), and placed on an indium tin oxide (ITO) coated slide (Delta Technologies, Auburn Hills, MI). Subsequently, a 1,5- diaminonaphthalene (DAN, Tokyo Chemical Industry Company Ltd., Tokyo, Cat. D0101) matrix (20 mg / mL in tetrahydrofuran) was applied to the tissue using an HTX M3+ sprayer (HTX Technologies, LLC, Chapel Hill, NC) to enhance laser energy absorption and ionization of tissue molecules for mass spectrometry. The spraying conditions were set to a nozzle temperature of 40 °C, nitrogen gas pressure of 15 psi, a solvent flow rate of 50 pL / min, and a track spacing of 2 mm, with a total of five passes over the tissue. Analysis of the tissue was conducted with a timsTOF fleX dual source mass spectrometer (Bruker Scientific, Billerica, MA), operated in positive ion mode. The analysis used a raster width of 100 pm . To ionize the tissue molecules, 150 laser shots were fired in a single burst. The instrument settings included a global attenuator value of 0%, a local laser power of 88%, and a m / z range of 50-2000. Data analysis for lipid annotation was carried out using SCiLS Lab version 2023 software (Bruker Scientific). Lipid putative identification consisted of errors less than 5 ppm and was achieved using LIPID MAPS database-

[0161] Xenium. Fresh-frozen samples were sliced into 10-micrometer-thick sections and placed onto the Xenium slides. Sections were digested to make the mRNA accessible. The following reagents were added: control probes for non-specific binding assessment, genomic DNA controls for signal source assessment, and 313 probes with two target RNA complementary sequences and one gene-specific barcode. Probes with a concentration of 10 nM were hybridized to the RNA at 50°C overnight. The tissue was washed to remove the unhybridized probes. Hybridized probes were ligated at 37°C for two hours. The primers for rolling circle amplification (RCA) were added to prepare for probe amplification. The probes were amplified in a two-step enzymatic process — first for one hour at 4°C and then for twohours at 37 °C. Multiple copies of the gene-specific barcodes were created, increasing the signal- to-noise ratio. A second wash was performed to chemically quench the background fluorescence. The tissue sections were placed into an imaging cassette and loaded into the Xenium Analyzer for image acquisition. Afterward, fluorescent oligonucleotides were added to the sample. During the 15 cycles of probe hybridization and imaging, Z-stacks images were captured at 0.75 pm intervals for each cycle. A spatial map of the RNA transcripts was created by stitching the Z- stacks images together, using the DAPI image as a reference. The decoding process involved using a Xenium codebook, which maps codewords to genes based on specific patterns of fluorescent signals across channels and cycles. These signals were compared against the codebook using a global maximum likelihood approach. A Q-Score was assigned to each decoded transcript to ensure accuracy, reflecting the confidence in its identity, with only those scoring 20 or higher included in downstream analyses. Control mechanisms, including negative control codewords, negative control probes, and unassigned codewords, were used to assess and calibrate the specificity and reliability of the decoding algorithm. Cell segmentation was performed using DAPI images to identify cell nuclei and Cellpose.

[0162] Stereo-seq. This protocol consists of several steps: tissue preparation and imaging, permeabilization and reverse transcription, cDNA purification and amplification, library construction, sequencing, and data analysis. Specifically, slides were rinsed twice with nuclease-free water and air-dried. Afterward, tissue sections were mounted onto the slides, placed on a heating plate, and incubated at 37°C for 5 minutes. The slides were subsequently transferred to methanol, pre-cooled to -20°C, and incubated for 30 minutes. The slides were then kept inside a fume hood to allow the methanol to evaporate for 5 minutes. Slides were transferred to a clean Petri dish, and 100 pL of tissue fluorescent staining solution was applied. This solution was composed of 94.5 pL 5X SSC, 0.5 pL qubit ssDNA reagent, and 5 pL Romanowsky-type I (RI). The slides were kept in the dark at room temperature for 5 minutes and rinsed with 100 pL of wash buffer. A coverslip was applied, and imaging was performed using a 10X objective lens on a FITC channel. After imaging, the coverslip was removed, and the slide was briefly immersed in 0. IX SSC for 5 seconds. Once dried, a cassette and gasket were placed on the slide. Next, 150 pL of IX permeabilization reagent solution was added to the chip, covered with sealing tape, and incubated at 37°C for 18 minutes. After incubation, the buffer was removed and followed by a rinse with 100 pL of PR rinse buffer. Reverse transcription was conducted using a mix containing reverse transcriptase enzyme, RT primers, and additives to synthesize complementary DNA (cDNA). The reaction was carried out at 42°C to ensure efficienttranscription. Following this step, cDNA release was achieved by incubating the tissue sections with a release buffer and collecting the synthesized cDNA into Eppendorf tubes for further processing. Purified cDNA was subjected to amplification to ensure sufficient yield for library preparation. Beads were used to purify the cDNA, removing contaminants and enhancing concentration. Amplification was performed using PCR, generating high-quality cDNA products suitable for downstream applications. Purified cDNA was quantified using Qubit and bioanalyzer systems to ensure precise measurements for library construction. Library construction involved cDNA fragmentation and barcode tagging to enable sample multiplexing. Fragmented cDNA was mixed with a PCR barcode primer mix and amplified under controlled conditions. Magnetic bead purification was employed to remove byproducts and ensure high library quality. The final library product was eluted in buffer, quantified, and evaluated for quality. Sequencing was performed using the DNBSEQ-T7 platform, followed by data analysis using the Stereo-seq Analysis Workflow (SAW) and data visualization using StereoMap.

[0163] Cell segmentation for Xenium and Stereo-seq. Cell segmentation for Xenium is achieved through a combination of DAPI staining and a deep neural network, with mRNA transcripts assigned to the nearest nucleus within a maximum distance of 15 pm. For Stereo-seq, segmentation is performed using a nucleic acid staining image and the Scikit-image package (v0.18.1).

[0164] Creation of an MALDI MSI image using a CSV file exported fromSCiLS Lab. The pipeline begins by parsing the CSV file to extract rows of interest and counting delimiters (semicolons) to assess data structure. Specific rows were extracted, and numeric values were parsed based on defined positional indices. For quality control, numeric values were filtered to ensure they fell within a target range (e.g., 885.5360 ± 0.0088). Spot-wise summation was performed for selected regions across multiple rows, generating a dictionary of summed values associated with each spot. A pixelated heatmap was generated using a custom matplotlib script, representing each spot as a red pixel with intensity proportional to its original intensity data. The resulting figure was saved as a high-resolution PNG image for downstream analysis.

[0165] Segmentation of MSI images using transformations and quantification of signal intensity. The cell boundary input image was upscaled by a factor of 4 using Lanczos interpolation to ensure label clarity, followed by Gaussian blurring of each RGB channel to enhance edge smoothness. Boundaries were identified using Canny edge detection and refined with morphological operations to close gaps. Contours were extracted, filtered by area size, and simplified using the Ramer-Douglas-Peucker algorithm. For each contour, themean RGB values were computed, and colors were classified by matching reference colors based on Euclidean distance in RGB space. Labeled outlines were created by overlaying identified contours with centroid labels, written as short text strings, on a transparent RGBA image. For image alignment, B-spline transformation or affine transformation can be employed. B-spline transformation operates by overlaying control points and computing displacement vectors for each point. These displacements are then interpolated using B-spline basis functions to create a continuous deformation field across the entire image. It can be mathematically represented as follows: Let T (x) be the transformation that maps a point x = (x, y, z) in the source image to its corresponding point in the target image. The B-spine transformation is defined as T(x) are the B-splinecoefficients (control points), / ?nis the n-th order B-spline basis function, nx, ny, nzare the control point spacings in each dimension.

[0166] Control point displacements between user-selected landmarks in both modalities were calculated and interpolated across a mesh grid using inverse distance weighting. The registered image was generated by resampling the input image using the updated B-spline parameters. The final output combined the registered images, edge overlays, and signal labels, upscaled further for improved resolution. Average signal intensity, area size, and total signal intensity were computed and saved in a .txt file. The pipeline for affine transformation is similar to that of B-spline transformation, except that affine transformation typically includes a series of operations such as translation T(x, y) = (x + tx, y + ty), rotation R(x, y) = (x cos 3 - y sin 9, x sin 9 + y cos 9 x), and scaling S(x, y) = (sxx„ syy) and maintains the parallelism of lines and the ratio of distances between points.

[0167] Segmentation of metabolic data using a rigid two-point approach.High-resolution images were exported from Xenium Explorer, with cell type information denoted by unique boundary colors. While exporting, zooming in on specific regions of interest is necessary for precise cell boundaries, depending on tissue size. Next, MALDI MSI data were visualized in SCiLS Lab, followed by importing the Xenium image into SCiLS Lab using the "Import Optical Image" function under the "File" tab. The metabolomics image was selected as the reference image. Its orientation was adjusted as needed. Two anchor points were selected on the Xenium image and labeled A and B to align the Xenium image with the metabolomics image. These points were then matched with corresponding shapes on the metabolomics image by dragging the anchors to the same locations. The co-registration was further adjusted byexamining the alignment across the tissue structure and texture. After completing image coregistration, the transparency of the metabolomics image was adjusted under “Ion Images” in “Visualization” until the boundary and color were distinguishable. Subsequently, the segmented cell regions were assigned using the “Polygon” tool. Each polygon was labeled with its corresponding cell type. Once all the cells of interest were selected and labeled, they were assigned into region groups based on their cell types using the “Create region with selected regions” function in “Regions.” These regions were analyzed in the “Feature Table,” where the average ion intensity of each segmented cellular region and region group was calculated. The analysis results were then visualized using the “Intensity Box Plot” and “ROC Plot.”

[0168] Although the subject matter has been described in language specific to structural features and / or methodological acts, it is to be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or acts described above. Rather, the specific features and acts described above are disclosed as example forms of implementing the claims.

[0169] Table 1 - Reagents and materials

[0170] References Pampols, T. (2010). Inherited Metabolic Rare Disease. Adv Exp Med Biol 686, 397-431. 10.1007 / 978-90-481-9485-8_23. Wang, M.M., Li, F.Q., Zhang, J., Lu, C., and Kong, W.J. (2023). Global Epidemiology of Gaucher Disease: an Updated Systematic Review and Meta- analysis. J Pediat Hematol One 45, 181-188. 10.1097 / Mph.0000000000002506. van Dussen, L., Biegstraaten, M., Hollak, C.E.M., and Dijkgraaf, M.G.W. (2014). Costeffectiveness of enzyme replacement therapy for type 1 Gaucher disease. Orphanet Journal of Rare Diseases 9. Artn 51 10.1186 / 1750-1172-9-51. Scriver, C.R. (2007). The PAH gene, phenylketonuria, and a paradigm shift. Hum Mutat 28, 831-845. 10.1002 / humu.20526. Rose, A.M., Grosse, S.D., Garcia, S.P., Bach, J., Kleyn, M., Simon, N.J.E., and Prosser, L.A. (2019). The financial and time burden associated with phenylketonuria treatment in the United States. Mol Genet Metab Rep 21. ARTN 10052310.1016 / j .ymgmr.2019.100523. Saeedi, P., Petersohn, I., Salpea, P., Malanda, B., Karuranga, S., Unwin, N., Colagiuri, S., Guariguata, L., Motala, A.A., Ogurtsova, K., et al. (2019). Global and regional diabetes prevalence estimates for 2019 and projections for 2030 and 2045: Results from the International Diabetes Federation Diabetes Atlas, 9th edition. Diabetes Res Clin Pr 157. ARTN 107843 10.1016 / j.diabres.2019.107843. Spraggins, J.M., Djambazova, K.V., Rivera, E.S., Migas, L.G., Neumann, E.K., Fuetterer, A., Suetering, J., Goedecke, N., Ly, A., Van de Plas, R., and Caprioli, R.M. (2019). High- Performance Molecular Imaging with MALDI Trapped Ion- Mobility Time-of- Flight (timsTOF) Mass Spectrometry. Analytical Chemistry 91, 14552-14560.10. 1021 / acs.analchem.9b03612. Ma, X., and Fernandez, F.M. (2024). Advances in mass spectrometry imaging for spatial cancer metabolomics. Mass Spectrom Rev 43, 235-268. 10.1002 / mas.21804. Zhu, X., Xu, T., Peng, C., and Wu, S. (2021). Advances in MALDI Mass Spectrometry Imaging Single Cell and Tissues. Front Chem 9, 782432. 10.3389 / fchem.2021.782432. Jiang, L.X., Yang, M., Wali, S.N., and Laskin, J. (2023). High-Throughput Mass Spectrometry Imaging of Biological Systems: Current Approaches and Future Directions. Trends Analyt Chem 163. 10.1016 / j.trac.2023.117055. Kompauer, M., Heiles, S., and Spengler, B. (2017). Atmospheric pressure MALDI massspectrometry imaging of tissues and cells at 1.4-pm lateral resolution. Nature Methods 14, 90-96. 10.1038 / Nmeth.4071. Zavalin, A., Yang, J., Hayden, K., Vestal, M., and Caprioli, R.M. (2015). Tissue protein imaging at 1 pm laser spot diameter for high spatial resolution and high imaging speed using transmission geometry MALDI TOF MS. Anal Bioanal Chem 407, 2337-2342. 10.1007 / s00216-015-8532-6. Scupakova, K., Dewez, F., Walch, A.K., Heeren, R.M.A., and Balluff, B. (2020). Morphometric Cell Classification for Single-Cell MALDI-Mass Spectrometry Imaging. Angew Chem Int Ed Engl 59, 17447-17450. 10.1002 / anie.202007315. Huber, D., von Voithenberg, L.V., and Kaigala, G. V. (2018). Fluorescence hybridization (FISH): History, limitations and what to expect from micro-scale FISH? Micro Nano Eng 1, 15-24. 10.1016 / j.mne.2018.10.006. He, S.S., Bhatt, R., Brown, C., Brown, E.A., Buhr, D.L., Chantranuvatana, K., Danaher, P., Dunaway, D., Garrison, R.G., Geiss, G., et al. (2022). High-plex imaging of RNA and proteins at subcellular resolution in fixed tissue by spatial molecular imaging. Nature Biotechnology 40. 10.1038 / s41587-022-01483-z. Janesick, A., Shelansky, R., Gottscho, A.D., Wagner, F., Williams, S.R., Rouault, M., Beliakoff, G., Morrison, C.A., Oliveira, M.F., Sicherman, J.T., et al. (2023). High resolution mapping of the tumor microenvironment using integrated single- cell, spatial and in situ analysis. Nature Communications 14. ARTN 8353 10.1038 / s41467-023- 43458-x. Chen, A., Liao, S., Cheng, M„ Ma, K„ Wu, L„ Lai, Y„ Qiu, X., Yang, J., Xu, J., Hao, S., et al. (2022). Spatiotemporal transcriptomic atlas of mouse organogenesis using DNA nanoball-pattemed arrays. Cell 185, 1777-1792 el721. 10.1016 / j.cell.2022.04.003. Stringer, C., Wang, T., Michaelos, M., and Pachitariu, M. (2021). Cellpose: a generalist algorithm for cellular segmentation. Nature Methods 18, 100-+. 10.1038 / s41592-020- 01018-x. Pachitariu, M., and Stringer, C. (2022). Cellpose 2.0: how to train your own model. Nature Methods 19, 1634-+. 10.1038 / s41592-022-01663-4. Arthur, D., and Vassilvitskii, S. (2007). k-means plus plus : The Advantages of Careful Seeding. Proceedings of the Eighteenth Annual Acm-Siam Symposium on Discrete Algorithms, 1027-1035. Zhang, Z.H., Murtagh, F., Van Poucke, S., Lin, S., and Lan, P. (2017). Hierarchicalcluster analysis in clinical research with heterogeneous study population: highlighting its visualization with R. Ann Transl Med 5. ARTN 75 10.21037 / atm.2017.02.05. Blondel, V.D., Guillaume, J.L., Lambiotte, R., and Lefebvre, E. (2008). Fast unfolding of communities in large networks. J Stat Meeh- Theory E. Artn P10008 10.1088 / 1742- 5468 / 2008 / 10 / P10008. Traag, V.A., Waltman, L., and van Eck, N.J. (2019). From Louvain to Leiden: guaranteeing well-connected communities. Sci Rep-Uk 9. ARTN 5233 10.1038 / s41598- 019-41695-z. Mackiewicz, A., and Ratajczak, W. (1993). Principal Components-Analysis (Pea). Comput Geosci 19, 303-342. Doi 10.1016 / 0098-3004(93)90090-R. van der Maaten, L., and Hinton, G. (2008). Visualizing Data using t-SNE. J Mach Learn Res 9, 2579-2605. Janesick, A., Shelansky, R., Gottscho, A.D., Wagner, F., Williams, S.R., Rouault, M., Beliakoff, G., Morrison, C.A., Oliveira, M.F., Sicherman, J.T., et al. (2023). High resolution mapping of the tumor microenvironment using integrated single-cell, spatial and in situ analysis. Nat Commun 14, 8353. 10.1038 / s41467-023- 43458-x. Traka, M., Wollmann, R.L., Cerda, S.R., Dugas, J., Barres, B.A., and Popko, B. (2008). Is a Nonsense Mutation in the Mouse Aspartoacylase Gene That Causes Spongy Degeneration of the CNS. Journal of Neuroscience 28, 11537-11549.10.1523 / Jneurosci.1490-08.2008. Patterson, N.H., Tuck, M., Van de Plas, R., and Caprioli, R.M. (2018). Advanced Registration and Analysis of MALDI Imaging Mass Spectrometry Measurements through Autofluorescence Microscopy. Analytical Chemistry 90, 12395-12403.10.102 l / acs. analchem.8b02884. Rappez, L., Stadler, M., Triana, S., Gathungu, R.M., Ovchinnikova, K., Phapale, P., Heikenwalder, M., and Alexandrov, T. (2021). SpaceM reveals metabolic states of single cells. Nat Methods 18, 799-805. 10.1038 / s41592-021 -01198-0. Bien, T., Koerfer, K., Schwenzfeier, J., Dreisewerd, K., and Soltwisch, J. (2022). Mass spectrometry imaging to explore molecular heterogeneity in cell culture. Proc Natl Acad Sci U S A 119, e2114365119. 10.1073 / pnas.2114365119. Podbielska, M., Ariga, T., and Pokryszko-Dragan, A. (2022). Sphingolipid Players in Multiple Sclerosis: Their Influence on the Initiation and Course of the Disease.International Journal of Molecular Sciences 23. ARTN 5330 10.3390 / ijms23105330.Maganti, R.J., Hronowski, X.P.L., Dunstan, R.W., Wipke, B.T., Zhang, X.L., Jandreski, L., Hamann, S., and Juhasz, P. (2019). Defining Changes in the Spatial Distribution and Composition of Brain Lipids in the Shiverer and Cuprizone Mouse Models of Myelin Disease. J Histochem Cytochem <57, 203-219. 10.1369 / 0022155418815860. Sen, M.K., Mahns, D.A., Coorssen, J.R., and Shortland, P.J. (2022). The roles of microglia and astrocytes in phagocytosis and myelination: Insights from the cuprizone model of multiple sclerosis. Glia 70, 1215-1250. 10.1002 / glia.24148. Filippatou, A.G., Moniruzzaman, M., Sotirchos, E.S., Fitzgerald, K.C., Kalaitzidis, G., Lambe, J., Vasileiou, E., Saidha, S., Prince, J.L., Haughey, N., et al. (2021). Serum ceramide levels are altered in multiple sclerosis. Mult Scler J 27, 1506- 1519. Artn1352458520971816 10.1177 / 1352458520971816. Podbielska, M., Szulc, Z.M., Ariga, T., Pokryszko-Dragan, A., Fortuna, W., Bilinska, M., Podemski, R., Jaskiewicz, E., Kurowska, E., Yu, R.K., and Hogan, E.L. (2020). Distinctive sphingolipid patterns in chronic multiple sclerosis lesions. J Lipid Res 61, 1464-1479. 10. 1194 / jlr.RA120001022. Papuc, E., and Rejdak, K. (2020). The role of myelin damage in Alzheimer's disease pathology. Arch Med Sci 16, 345-351. 10.5114 / aoms.2018.76863. Kaya, L, Nilsson, A., Luptakova, D., He, Y., Vallianatou, T., Bjarterot, P., Svenningsson, P., Bezard, E., and Andren, P.E. (2023). Spatial lipidomics reveals brain region- specific changes of sulfatides in an experimental MPTP Parkinson’s disease primate model. NPJ Parkinsons Dis 9, 118. 10.1038 / s41531-023-00558- 1. Xu, S.L., Zhu, Z.J., Delafield, D.G., Rigby, M.J., Lu, G.Y., Braun, M„ Puglielli, L„ and Li, L.J. (2024). Spatially and temporally probing distinctive glycerophospholipid alterations in Alzheimer's disease mouse brain via high- resolution ion mobility -enabled- position resolved lipidomics. Nature Communications 15. ARTN 6252 10.1038 / s41467- 024-50299-9. Momchilova, A., Pankov, R., Alexandrov, A., Markovska, T., Pankov, S., Krastev, P., Staneva, G., Vassileva, E., Krastev, N., and Pinkas, A. (2022). Sphingolipid Catabolism and Glycerophospholipid Levels Are Altered in Erythrocytes and Plasma from Multiple Sclerosis Patients. International Journal of Molecular Sciences 23. ARTN 7592 10.3390 / ijms23147592. Stoessel, D., Stellmann, J.P., Willing, A., Behrens, B., Rosenkranz, S.C., Hodecker, S.C.,Sturner, K.H., Reinhardt, S., Fleischer, S., Deuschle, C., et al. (2018). MetabolomicProfiles for Primary Progressive Multiple Sclerosis Stratification and Disease Course Monitoring. Front Hum Neurosci 12. ARTN 226 10.3389 / fnhum.2018.00226. Fonteh, A.N., Chiang, J.R., Cipolla, M., Hale, J., Diallo, F., Chirino, A., Arakaki, X., and Harrington, M.G. (2013). Alterations in cerebrospinal fluid glycerophospholipids and phospholipase A activity in Alzheimer's disease. J Lipid Res 54, 2884-2897.10.1194 / jlr.M037622. Gudi, V., Grieb, P., Linker, R.A., and Skripuletz, T. (2023). CDP-choline to promote remyelination in multiple sclerosis: the need for a clinical trial. Neural Regen Res 18, 2599-2605. 10.4103 / 1673-5374.373671. Wheeler, D., Bandaru, V.V.R., Calabresi, P.A., Nath, A., and Haughey, N.J. (2008). A defect of sphingolipid metabolism modifies the properties of normal appearing white matter in multiple sclerosis. Brain 131, 3092-3102. 10.1093 / brain / awnl90. Villoslada, P., Alonso, C., Agirrezabal, I., Kotelnikova, E., Zubizarreta, L, Pulido- Valdeolivas, I., Saiz, A., Comabella, M., Montalban, X., Villar, L., et al. (2017). Metabolomic signatures associated with disease severity in multiple sclerosis. Neurol Neuroimmunol Neuroinflamm 4, e321. 10.1212 / NXI.0000000000000321. Bura, A., Cabrijan, S., Duric, I., Bruketa, T., and Begonja, A.J. (2023). A Plethora of Functions Condensed into Tiny Phospholipids: The Story of PI4P and PI(4,5)P. Celis- Basel 12. ARTN 1411 10.3390 / cellsl2101411. Dickson, E.J. (2019). Recent advances in understanding phosphoinositide signaling in the nervous system. FlOOORes 8. 10.12688 / fl000research.16679.1. Denti, V., Capitoli, G., Piga, I., Clerici, F., Pagani, L., Criscuolo, L., Bindi, G., Principi, L., Chinello, C., Paglia, G., et al. (2022). Spatial Multiomics of Lipids, N- Glycans, and Tryptic Peptides on a Single FFPE Tissue Section. J Proteome Res. 10.1021 / acs.jproteome.2c00601. Ngai, Y.T., Briggs, M.T., Mittal, P., Young, C., Parkinson-Lawrence, E., Klingler- Hoffmann, M., Orgeig, S., and Hoffmann, P. (2024). Mass spectrometry imaging protocol for spatial mapping of lipids, -glycans and peptides in murine lung tissue. Rapid Commun Mass Sp 38. ARTN e9721 10.1002 / rcm.972L Carter, C.L., Parker, G.A., Hankey, K.G., Farese, A.M., MacVittie, T.J., and Kane, M.A. (2020). MALDI-MSI spatially maps N-glycan alterations to histologically distinct pulmonary pathologies following irradiation. Sci Rep-Uk 10. ARTN 11559 10.1038 / s41598-020-68508-y.Wei, X., Fu, S., Li, H„ Liu, Y„ Wang, S., Feng, W., Yang, Y„ Liu, X., Zeng, Y.Y., Cheng, M., et al. (2022). Single-cell Stereo-seq reveals induced progenitor cells involved in axolotl brain regeneration. Science 577, eabp9444. 10.1126 / science.abp9444. Kinkhabwala, A., Herbel, C., Pankratz, J., Yushchenko, D.A., Ruberg, S., Praveen, P., Reiss, S., Rodriguez, F.C., Schafer, D., Kollet, J., et al. (2022). MACSima imaging cyclic staining (MICS) technology reveals combinatorial target pairs for CAR T cell treatment of solid tumors. Sci Rep-Uk 12. ARTN 1911 10.1038 / s41598-022-05841-4. Conroy, M.J., Andrews, R.M., Andrews, S., Cockayne, L., Dennis, E.A., Fahy, E., Gaud, C„ Griffiths, W.J., Jukes, G„ Kolchin, M„ et al. (2024). LIPID MAPS: update to databases and tools for the lipidomics community. Nucleic Acids Res 52, D1677-D1682. 10.1093 / nar / gkad896. Chen, A., Liao, S., Cheng, M.N., Ma, K.L., Wu, L„ Lai, Y.W., Qiu, X.J., Yang, J., Xu, J.S., Hao, S.J., et al. (2022). Spatiotemporal transcriptomic atlas of mouse organogenesis using DNA nanoball-pattemed arrays. Cell 185, 1777-+. 10.1016 / j.cell.2022.04.003. van der Walt, S., Schonberger, J.L., Nunez-Iglesias, J., Boulogne, F., Warner, J.D., Yager, N., Gouillart, E., Yu, T., and Contributors, S.I. (2014). scikit-image: image processing in Python. Peer) 2. ARTN e453 10.7717 / peerj.453. Sud, M., Fahy, E., Cotter, D., Brown, A., Dennis, E.A., Glass, C.K., Merrill, A.H., Murphy, R.C., Raetz, C.R.H., Russell, D.W., and Subramaniam, S. (2007). LMSD: LIPID MAPS structure database. Nucleic Acids Research 55, D527-D532. 10.1093 / nar / gkl838.

Claims

WHAT IS CLAIMED:

1. A computer-implemented method of single-cell spatial omics analysis comprising: receiving a plurality of images that capture at least one cell having a complex shape, the plurality of images comprising a first stained image and a second stained image, wherein the first stained image enhances the visibility or contrast of the at least one cell’s nucleus, and wherein the second stained image enhances the visibility or contrast of the at least one cell’s edges; receiving a data matrix that represents a gene expression profile of the at least one cell; combining the first stained image and the second stained image to create an overlay image; analyzing at least one of the second stained image or the overlay image to delineate a boundary of the at least one cell; creating a mask based on the boundary of the at least one cell in the at least one of the second stained image or the overlay image; and combining the mask and the data matrix to form a combined image for the at least one cell.

2. The computer-implemented method of claim 1, further comprising performing an omics analysis of the combined image.

3. The computer-implemented method of claim 2, wherein the omics analysis comprises determining transcript counts for the at least one cell.

4. The computer-implemented method of claim 2 or 3, wherein the omics analysis comprises determining a spatial distribution of the gene expression profile of the at least one cell.

5. The computer-implemented method of any one of claims 2-4, wherein the omics analysis is at least one of a cell typing analysis, a cell-cell communication analysis, an analysis of the number and distribution of different cell types, a differential gene expression analysis, a marker gene identification analysis, a protein-protein interaction network analysis, a cell trajectory analysis, or a functional enrichment analysis.

6. The computer-implemented method of any one of claims 1-5, wherein the step of analyzing at least one of the second stained image or the overlay image to delineate the boundary of the at least one cell comprises: filtering the at least one of the second stained image or the overlay image based on at least one of a signal intensity or a structure size; performing an edge detection process to identify a plurality of pixels that form the boundary of the at least one cell; retaining the plurality of pixels that form the boundary of the at least one cell, wherein the boundary of the at least one cell includes a plurality of gaps between pixels of the plurality of pixels; and connecting the boundary of the at least one cell by extending through the plurality of gaps.

7. The computer-implemented method of any one of claims 1-6, further comprising: receiving a mass spectrometry image that captures the at least one cell, wherein the mass spectrometry image is a spatial representation of a plurality of lipids or metabolites within the at least one cell; combining information associated with the combined image and the mass spectrometry image to form an integrated image; and analyzing the integrated image to determine a metabolic profile of the at least one cell.

8. The computer-implemented method of claim 7, wherein the information associated with the combined image is the boundary of the at least one cell and a type of the at least one cell.

9. The computer-implemented method of claim 7 or 8, further comprising determining a signal intensity and a location of at least one metabolite and a metabolic state of the at least one cell.

10. The computer-implemented method of any one of claims 1-9, wherein the at least one cell is elongated, cylindrical, polygonal spindle shaped, pyramid shaped, dome shaped, or star shaped.

11. The computer-implemented method of any one of claims 1-9, wherein the at least one cell comprises a plurality of branching structures.

12. The computer-implemented method of any one of claims 1-9, wherein the complex shape is not a round or elliptical shape.

13. The computer-implemented method of any one of claims 1-9, wherein the at least one cell is a neuronal cell, an astrocyte, a fibroblast, a mesenchymal stem cell, a smooth muscle cell, a skeletal muscle fiber, an osteoclast, a keratinocyte, an endothelial cell, or a Sertoli cell.

14. The computer-implemented method of any one of claims 1-13, wherein the at least one cell is stained with a dye that binds to deoxyribonucleic acid (DNA) in the first stained image.

15. The computer-implemented method of any one of claims 1-14, wherein the at least one cell is stained with an antibody in the second stained image.

16. A method comprising: performing, for a biological sample from a subject, the computer-implemented method of single-cell spatial omics analysis according to any one of claims 1-15; and determining a diagnosis, prognosis, or treatment plan for the subject based on one or more respective cell-by-gene matrices for one or more cells of the biological sample.

17. The method of claim 16, further comprising administering a treatment to the subject based on the diagnosis, prognosis, or treatment plan.

18. A computer-implemented method comprising: receiving a plurality of images that capture at least one cell having a complex shape, the plurality of images comprising a first stained image and a second stained image, wherein the first stained image enhances the visibility or contrast of the at least one cell’s nucleus, and wherein the second stained image enhances the visibility or contrast of the at least one cell’s edges; combining the first stained image and the second stained image to create an overlay image;analyzing at least one of the second stained image or the overlay image to delineate a boundary of the at least one cell; and segmenting the at least one cell based on the boundary of the at least one cell in the at least one of the second stained image or the overlay image.

19. A computer-implemented method of single-cell spatial omics analysis comprising: receiving a plurality of images that capture at least one cell, the plurality of images comprising a first stained image and a second stained image, wherein the first stained image enhances the visibility or contrast of the at least one cell’s nucleus, and wherein the second stained image enhances the visibility or contrast of the at least one cell’s edges; receiving a data matrix that represents a gene expression profile of the at least one cell; combining the first stained image and the second stained image to create an overlay image; analyzing at least one of the second stained image or the overlay image to delineate a boundary of the at least one cell; creating a mask based on the boundary of the at least one cell in the at least one of the second stained image or the overlay image; and combining the mask and the data matrix to form a combined image for the at least one cell, wherein the step of analyzing at least one of the second stained image or the overlay image to delineate the boundary of the at least one cell comprises: filtering the at least one of the second stained image or the overlay image based on at least one of a signal intensity or a structure size; performing an edge detection process to identify a plurality of pixels that form the boundary of the at least one cell; retaining the plurality of pixels that form the boundary of the at least one cell, wherein the boundary of the at least one cell includes a plurality of gaps between pixels of the plurality of pixels; and connecting the boundary of the at least one cell by extending through the plurality of gaps.

20. A computer-implemented method of single-cell spatial metabolomics analysis comprising:receiving a transcriptomic image for at least one cell, wherein the transcriptomic image is a spatial representation of a gene expression profile of the at least one cell; receiving a mass spectrometry image that captures the at least one cell, wherein the mass spectrometry image is a spatial representation of a plurality of lipids or metabolites within the at least one cell; combining information associated with the transcriptomic image and the mass spectrometry image to form an integrated image; and analyzing the integrated image to determine a metabolic profile of the at least one cell.

21. The computer-implemented method of claim 20, wherein the information associated with the transcriptomic image is a boundary of at least one cell and a type of the at least one cell.

22. The computer-implemented method of claim 20 or 21 , further comprising determining a signal intensity and a location of at least one metabolite and a metabolic state of the at least one cell.

23. The computer-implemented method of any one of claims 20-22, further comprising generating display data for the integrated image.

24. The computer-implemented method of any one of claims 20-23, wherein the transcriptomic image for the at least one cell is obtained by performing a single-cell spatial omics analysis.

25. A method comprising: performing, for a biological sample from a subject, the computer-implemented method of single-cell spatial metabolomics analysis according to any one of claims 20-24; and determining a diagnosis, prognosis, or treatment plan for the subject based on one or more respective metabolic profiles for one or more cells of the biological sample.

26. The method of claim 25, further comprising administering a treatment to the subject based on the diagnosis, prognosis, or treatment plan.

27. A method comprising: seeding a plurality of cells on a surface; treating the plurality of cells with one or more therapeutic agents; performing, for a biological sample from the plurality of cells, the computer-implemented method of single-cell spatial omics analysis according to any one of claims 1-15; and determining a therapeutic effect of the one or more therapeutic agents based on the single-cell spatial omics analysis.

28. A method comprising: seeding a plurality of cells on a surface; treating the plurality of cells with one or more therapeutic agents; performing, for a biological sample from the plurality of cells, the computer-implemented method of single-cell spatial metabolomics analysis according to any one of claims 20-24; and determining a therapeutic effect of the one or more therapeutic agents based on the single-cell spatial metabolomics analysis.

Images