Cell type identification

Abstract

Provided herein are compositions and methods for identifying the cell type origin of cells based on RNA sequencing (RNA-seq) data.The method provided herein may include: receiving a plurality of sequencing read counts; providing a set of detected cell types and a related gene set G for each cell type, such that each gene set G comprises a plurality of genes g; scoring the sequencing read counts to generate an assignment score; and assigning each cell among the plurality of cells to the cell type with the highest assignment score, thereby identifying the cell type origin of each cell.This method can be implemented, for example, using a computer system.

Description

[Technical Field]

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS This application claims the benefit of U.S. Provisional Patent Application No. 63 / 428,267, filed November 28, 2022, and European Patent Application No. 23315048.1, filed March 10, 2023, the contents of which are incorporated herein by reference in their entireties.

[0002] The present disclosure relates to identifying cell types based on RNA sequencing (RNA-seq) data. [Background technology]

[0003] Single-cell RNA sequencing (scRNA-seq) experiments offer the opportunity to observe complex tissues at single-cell resolution. However, insightful biological interpretation of scRNA-seq data relies on accurate identification of cell types. Rapid and accurate identification of cellular origin from scRNA-seq data can improve downstream analyses. However, cell type identification can be challenging due to variability in phenotype and cellular state.

[0004] Manual annotation using standard scRNA-seq data analysis workflows typically relies on manually examining the expression of cell-type-specific markers to identify cell types through reference marker genes and assign each group to a specific cell type. These marker features are either known from previous studies or identified using differential expression analysis of a given cell group relative to the rest of the dataset. However, this strategy is time-consuming, biased, and prone to error. Manual annotation requires preprocessing of high-dimensional data for visualization to the user. This preprocessing reshapes the data so that cells with similar biological patterns of transcripts have similar transformed measurements, thereby resizing them to be closer to each other in a reduced-dimensional gene expression space. This preprocessing step can introduce unintended distortions that affect downstream analyses. Additionally, clustering algorithms reduce the variability present in high-dimensional inputs, thereby attenuating biological signals. Finally, manual cell type assignment is user-dependent and susceptible to researcher bias. Summary of the Invention [Means for solving the problem]

[0005] This disclosure is based, in part, on the development of a cell-type annotation method that operates at individual cell resolution by applying a scoring system to scRNA-seq data based on a set of marker genes associated with the cell type. The method disclosed herein is transform- and cluster-free, avoiding undesirable distortions introduced by preprocessing steps. Disclosed herein are the foundations of the method, in silico experiments benchmarking its performance on simulated data, and a comparison of automated annotations by the method with manually annotated experimental data from the Tabula Sapiens Consortium (Tabula Sapiens Consortium. The Tabula Sapiens: A multiple-organ, single-cell transcriptomic atlas of humans. Science. 2022 May 13;376(6594).).

[0006] According to one aspect of the present disclosure, a method for identifying the cell type origin of each cell of a plurality of cells is provided. In one aspect, the method may include receiving a plurality of sequencing read counts; providing a set of detected cell types and an associated gene set G for each cell type, such that each gene set G includes a plurality of genes g; scoring the sequencing read counts to generate an assignment score; and assigning each cell of the plurality of cells to the cell type with the highest assignment score to identify the cell type origin of each cell. In some embodiments, the method further includes trimming the sequencing read counts to remove ambiguous calls. In some embodiments, the method further includes smoothing annotations of the sequencing read counts. In some embodiments, the plurality of sequencing read counts are generated by RNA sequencing (RNA-seq). In some embodiments, the plurality of sequencing read counts are generated by single-cell RNA sequencing (scRNA-seq) for each cell of the plurality of cells. In some embodiments, the scoring step includes sorting the plurality of genes g of the associated gene set G that exhibit non-zero expression from high to low expression. In some embodiments, the scoring step further comprises generating a ranked vector of length N from the sorted genes of the relevant gene sets. In some embodiments, the scoring step further comprises converting the ranked vector of length N into a binary array s such that genes g included in a particular gene set G are replaced by 1 or 0.

[0007] In some embodiments, the binary sequence s is represented by the following formula:

number

[0008] In some embodiments, the scoring step further comprises performing a cumulative sum over the binary array s for each element k of the binary vector, In some embodiments, the scoring step further comprises performing a sum over all generated successive partial sums from 1 to k to generate an assignment score S for the given cell according to the formula:

number

[0009] In some embodiments, the scoring step further comprises performing one or more of the foregoing steps for each cell and gene set of the plurality of cells, for a total of M gene sets. In some embodiments, the scoring step further comprises converting the input gene x cell expression matrix into a gene set x cell assignment score matrix, wherein the input gene x cell expression matrix comprises [number of genes x number of cells], and the gene set x cell assignment score matrix comprises [number of gene sets x number of cells].

[0010] In some embodiments, each cell of the plurality of cells has the formula: max{S m} 1≦m≦M· The cells are assigned to the cell type that represents the highest assignment score, represented by

[0011] In some embodiments, the trimming step further comprises calculating a Gini index value for each cell of the plurality of cells. In some embodiments, the trimming step calculates the gene set x cell assignment-score matrix using the following formula: ∈[0,1] and further converting the distribution to an exponential distribution according to

[0012] In some embodiments, the trimming step further comprises removing cells that are outliers and / or have an index less than 0.5. In some embodiments, the smoothing step further comprises applying a k-nearest neighbor (kNN) algorithm to the sequencing read counts.

[0013] In some embodiments, the smoothing step further comprises identifying cells with a minimum k nearest neighbors, and annotating the neighboring cell according to the consensus of the neighbors if more than 50% of the neighbors reach a consensus annotation. In some embodiments, the method further comprises providing a set of negative markers, wherein cells that do not express one or more negative markers are rewarded and cells that express one or more negative markers are penalized.

[0014] In some embodiments, the method does not include normalization of the plurality of sequencing read counts. In some embodiments, the method does not include transformation of the plurality of sequencing read counts. In some embodiments, the method does not include clustering of the plurality of sequencing read counts.

[0015] In some embodiments, the plurality of cells are human cells. In some embodiments, the set of cell types to be detected includes 1, 2, 3, 4, 5, 6, 7, 8, 9, or 10 or more cell types to be detected. In some embodiments, the plurality of genes g includes 10, 20, 30, 40, 50, 60, 70, 80, 90, or 100 or more genes.

[0016] In another aspect of the present disclosure, a computer program product tangibly embodied in a computer-readable medium is provided. In one aspect of the present disclosure, the computer program may include instructions configured, when executed by one or more processors, to receive a plurality of sequencing read counts, receive a set of detected cell types and an associated gene set G for each cell type, such that each gene set G includes a plurality of genes g, score the sequencing read counts to generate an assignment score, and assign each cell of the plurality of cells to the cell type with the highest assignment score to identify the tissue type origin of each cell. In some embodiments, the computer program further includes instructions configured, when executed by one or more processors, to trim the sequencing read counts to remove ambiguous calls. In some embodiments, the computer program further includes instructions configured, when executed by one or more processors, to smooth the annotation of the sequencing read counts.

[0017] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. Methods and materials are described herein for use in the present invention; other suitable methods and materials known in the art can also be used. The materials, methods, and examples are illustrative only and are not intended to be limiting. All publications, patent applications, patents, sequences, database entries, and other references mentioned herein are incorporated by reference in their entirety. In case of conflict, the present specification, including definitions, will control.

[0018] Other features and advantages of the methods and compositions will become apparent from the following detailed description and drawings, and from the claims. [Brief explanation of the drawings]

[0019] [Figure 1] 1 is a flow chart illustrating sequential steps of the method disclosed herein. [Figure 2] 1 is a flow chart further illustrating sequential steps of the methods disclosed herein. [Figure 3A] Schematic representation of the input of the method disclosed herein, where expression of a single gene (row) is assessed for each single cell (column) in an experiment based on scRNA-seq data. [Figure 3B] 1 is a schematic representation of the scoring step of the methods disclosed herein, where single cells (columns) are scored against a provided set of genes (rows). [Figure 3C] 1 is a schematic representation of the trimming step of the method disclosed herein, in which the Gini index is used as a measure of dispersion to identify ambiguous calls and unconfirmed unclassified cells for the remaining steps of the method. [Figure 3D] 1 is a schematic representation of the smoothing step of the method disclosed herein, in which cells with a minimum k nearest neighbors are identified by aggregating information from cells (neighbors) with similar expression profiles across the genome. [Figure 4A] 1 is a table showing the number of cells, genes, and cell types in the simulation data used to benchmark the cell type identification methods disclosed herein. [Figure 4B] FIG. 1 depicts the study design for benchmarking the cell type identification methods disclosed herein using simulated data including a training dataset, a marker gene set, and a test dataset. [Figure 4C] 1 is a set of three plots showing the benchmark performance of the cell type identification method disclosed herein using simulated data, showing sensitivity (left), specificity (center), and F1 score (right). [Figure 4D] 1 is a set of three plots showing benchmarking performance of the cell type identification method disclosed herein using simulated data for discovering novel cell types, where the simulated data corresponds to 5,000 cells, 10,000 cells, and 15,000 cells. Sensitivity (left), specificity (center), and F1 score (right) are shown. [Figure 5] FIG. 1 is a component diagram of a computer system that can be used to implement a method for identifying the cell type origin of each cell of a plurality of cells based on scRNA-seq data. [Figure 6A] Cluster plots show two-dimensional maps of cells organized based on similarity in gene expression profiles. Data from the Tabula Sapiens Consortium (left) was compared with annotations using the cell type identification method disclosed herein for peripheral blood mononuclear cells (PBMCs) (right). [Figure 6B] 1 is a plot of the Jaccard similarity index comparing Tabula Sapiens Consortium data with annotations using the cell type identification method disclosed herein for peripheral blood mononuclear cells (PBMCs). [Figure 6C] 1 shows a plot of PBMC marker gene expression by cell type (y-axis) and reference marker (x-axis) comparing Tabula Sapiens Consortium data with annotation using the cell type identification method disclosed herein. Circle size corresponds to the percentage of cells expressing the marker, and shading corresponds to expression level. [Figure 7A] A series of t-stochastic neighbor embedding (t-SNE) plots for each of the nine reference markers for T cells (TC) and natural killer cells (NK). [Figure 7B] Cluster plots show two-dimensional maps of cells organized based on similarity in gene expression profiles, comparing Tabula Sapiens Consortium data (top) with annotations for T cells (TC) and natural killer cells (NK) using the cell type identification method disclosed herein (bottom). [Figure 7C]1 is a plot of natural killer cell marker gene expression by cell type (y-axis) and reference marker (x-axis) comparing Tabula Sapiens Consortium data with annotation using the cell type identification method disclosed herein. Circle size corresponds to the percentage of cells expressing the marker, and shading corresponds to expression level. [Figure 7D] 1 is a plot of the Jaccard similarity index comparing Tabula Sapiens Consortium data with annotations using the cell type identification method disclosed herein for T cells (TC) and natural killer cells (NK). [Figure 7E] 1 is a plot of expression levels of reference T cell (TC) and natural killer (NK) cell markers for cell populations identified as NK cells by the cell type identification method disclosed herein and as T cells by the Tabula test. [Figure 8A] A series of t-stochastic neighbor embedding (t-SNE) plots for each of the four reference markers for T cells (TCs). [Figure 8B] Cluster plots showing two-dimensional maps of cells organized based on similarity in gene expression profiles, comparing Tabula Sapiens Consortium data (top) with annotations for T cells (TCs) using the cell type identification method disclosed herein (bottom). [Figure 8C] 1 shows a plot of T cell marker gene expression by cell type (y-axis) and reference marker (x-axis) comparing Tabula Sapiens Consortium data with annotation using the cell type identification method disclosed herein. Circle size corresponds to the percentage of cells expressing the marker, and shading corresponds to expression level. [Figure 8D] 1 is a plot of the Jaccard similarity index comparing Tabula Sapiens Consortium data with annotations using the cell type identification method disclosed herein for T cells (TCs). [Figure 8E] 1 is a plot of the expression levels of three T cell reference markers, CD4, CD8A, and CD8B, for mixed populations of cells identified as CD4+ by the cell typing methods disclosed herein and CD8+ by the Tabula test (left) or cells identified as CD8+ by the cell typing methods disclosed herein and CD4+ by the Tabula test (right). [Figure 9A] Cluster plots showing two-dimensional maps of cell organization based on similarity in gene expression profiles for macrophages (Mac) and monocytes (Mon) compared with Tabula Sapiens Consortium data (left) and annotations using the cell type identification method disclosed herein (right). [Figure 9B] 1 shows a plot of reference marker expression for macrophages and monocytes plotted by cell type (y-axis) and reference marker (x-axis) comparing Tabula Sapiens Consortium data with annotation using the cell type identification method disclosed herein. Circle size corresponds to the percentage of cells expressing the marker, and shading corresponds to expression level. [Figure 9C] 1 is a plot of the Jaccard similarity index comparing Tabula Sapiens Consortium data with annotations using the cell type identification method disclosed herein for macrophages (Mac) and monocytes (Mon). [Figure 9D] 1 is a series of heat maps of CD16 and CD14 expression on classical, intermediate, and non-classical monocytes annotated by the cell type identification method disclosed herein (top panel) and the Tabula test (bottom panel). [Figure 10A] Cluster plots show two-dimensional maps of cells organized based on similarity in gene expression profiles. Data from the Tabula Sapiens Consortium (left) was compared with annotations of naive B cells, memory B cells, and plasma cells using the cell type identification method disclosed herein (right). [Figure 10B] 1 shows a plot of reference marker expression for naive B cells (B.Naive), memory B cells (B.Mem), and plasma cells plotted by cell type (y-axis) and reference marker (x-axis), comparing Tabula Sapiens Consortium data with annotation using the cell type identification method disclosed herein. Circle size corresponds to the percentage of cells expressing the marker, and shading corresponds to expression level. [Figure 10C] 1 is a plot of the Jaccard similarity index comparing Tabula Sapiens Consortium data with annotations using the cell type identification methods disclosed herein for naive B cells (B.Naive), memory B cells (B.Mem), and plasma cells. [Figure 10D] 1 is a plot of the expression levels of B cell reference markers for a mixed population of cells identified as naive B cells by the cell typing methods disclosed herein and as memory B cells by the Tabula test. [Figure 11A] Cluster plots show two-dimensional maps of cells organized based on similarity in gene expression profiles. Data from the Tabula Sapiens Consortium (left) is compared with annotations using the cell type identification method disclosed herein (right) for cardiac tissue, including cardiac fibroblasts (CFs), cardiomyocytes (CMCs), endothelial cells (Endo), hepatocytes (Hepa), macrophages (Macs), and smooth muscle cells (SMCs). [Figure 11B] Figure 1 shows a plot of reference marker expression for cardiac fibroblasts (CFs), cardiomyocytes (CMCs), endothelial cells (Endo), hepatocytes (Hepa), macrophages (Macs), and smooth muscle cells (SMCs) plotted by cell type (y-axis) and reference marker (x-axis) comparing annotations using the cell type identification method disclosed herein with data from the Tabula Sapiens Consortium. The size of the circle corresponds to the percentage of cells expressing the marker, and the degree of shading corresponds to the expression level. [Figure 11C]Figure 1 shows a plot of the Jaccard similarity index for cardiac fibroblasts (CFs), cardiomyocytes (CMCs), endothelial cells (Endo), hepatocytes (Hepa), macrophages (Macs), and smooth muscle cells (SMCs) compared to annotations using the cell type identification method disclosed herein and data from the Tabula Sapiens Consortium. [Figure 12A] Cluster plots show two-dimensional maps of cells organized based on similarity in gene expression profiles, comparing the Tabula Sapiens Consortium data (left) with annotations of human kidney tissues (including B cells (B), endothelial cells (Endo), epithelial cells (Epit), macrophages (Mac), T cells, and natural killer cell populations (TNK)) using the cell type identification method disclosed herein (right). [Figure 12B] Figure 1 shows a comparison of annotations using the cell type identification method disclosed herein with data from the Tabula Sapiens Consortium, plotting the expression of reference markers for human kidney tissues, including B cells (B), endothelial cells (Endo), epithelial cells (Epit), macrophages (Mac), T, and natural killer cell populations (TNK), by cell type (y-axis) and reference marker (x-axis). The size of the circle corresponds to the percentage of cells expressing the marker, and the degree of shading corresponds to the expression level. [Figure 12C] FIG. 1 is a plot of the Jaccard similarity index comparing annotations using the cell type identification methods disclosed herein with data from the Tabula Sapiens Consortium for human kidney tissue cell types, including B cells (B), endothelial cells (Endo), epithelial cells (Epit), macrophages (Mac), T and natural killer cell populations (TNK). [Figure 13A]Cluster plots show two-dimensional maps of cells organized based on similarity in gene expression profiles. Data from the Tabula Sapiens Consortium (left) is compared with annotations (right) using the cell type identification method disclosed herein for human lung tissues, including adventitial cells (AC), alveolar type 1 and type 2 cells (ATC), basal cells (Basal), basophils (Baso), B cells and plasma cells (BPC), club cells (CC), endothelial cells (Endo), fibroblasts (Fibro), goblet cells, serous cells, mucus cells (GSM), pulmonary ciliated cells (LCC), mesothelial cells (MC), monocytes, dendritic cells, macrophages (MPh), neutrophils (Neut), pericytes (PC), lung ionocytes (PI), smooth muscle cells (SMC), T cells, and natural killer cells (TNK). [Figure 13B] Figure 1 shows a plot of the expression of reference markers for human lung tissue, including adventitial cells (AC), alveolar type 1 and type 2 cells (ATC), basal cells (Bas), basophils (Baso), B cells and plasma cells (BPC), club cells (CC), endothelial cells (Endo), fibroblasts (Fibro), goblet cells, serous cells and mucus cells (GSM), pulmonary ciliated cells (LCC), mesothelial cells (MC), monocytes, dendritic cells and macrophages (MPh), neutrophils (Neut), pericytes (PC), lung ionocytes (PI), smooth muscle cells (SMC), T cells and natural killer cells (TNK) reference markers, plotted against cell type (y-axis) and reference marker (x-axis) using annotations using the cell type identification methods disclosed herein compared with data from the Tabula Sapiens Consortium. The size of the circle corresponds to the percentage of cells expressing the marker, and the degree of shading corresponds to the expression level. [Figure 13C]1 is a plot of the Jaccard similarity index comparing data from the Tabula Sapiens Consortium with annotations using the cell type identification method disclosed herein for complex human lung tissue including adventitial cells (AC), alveolar type 1 and type 2 cells (ATC), basal cells (Basal), basophils (Baso), B cells and plasma cells (BPC), club cells (CC), endothelial cells (Endo), fibroblasts (Fibro), goblet cells, serous cells and mucous cells (GSM), pulmonary ciliated cells (LCC), mesothelial cells (MC), monocytes, dendritic cells and macrophages (MPh), neutrophils (Neut), pericytes (PC), lung ionocytes (PI), smooth muscle cells (SMC), T cells and natural killer cells (TNK). DETAILED DESCRIPTION OF THE INVENTION

[0020] Scoring 1 is a flowchart illustrating sequential steps of the methods disclosed herein. In some embodiments, the method initially includes a receiving step 102 in which a plurality of sequencing read counts are received by a computer-implemented system. In a providing step 104, a set of cell types to be detected and an associated gene set G for each cell type are provided by a user, such that each gene set G includes a plurality of genes g. In a scoring step 106, the sequencing read counts are scored to generate an assignment score, as described in more detail below. In step 108, each cell of the plurality of cells is assigned to the cell type with the highest assignment score to identify the tissue type origin of each cell.

[0021] FIG. 2 is a flowchart illustrating the sequential steps of the method disclosed herein in further detail. In some embodiments, the method initially includes receiving step 202, in which a plurality of single-cell RNA sequencing read counts are received by a computer-implemented system. In step 204, a user provides a cell type to be detected and an associated gene set G for each cell type, such that each gene set G includes a plurality of genes g. In step 206, the plurality of genes g in the associated gene set G that exhibit non-zero expression are sorted from high to low expression. In step 208, a ranked vector of length N is generated from the sorted genes in the associated gene set. In step 210, the ranked vector of length N is converted to a binary array s such that genes g included in a particular gene set G are replaced by either 1 or 0. In step 212, a cumulative sum is performed over the binary array s for each element k of the binary vector. In step 214, a sum is performed over all generated consecutive partial sums from 1 to k to generate an assignment score S for each cell. In step 216, each cell of the plurality of cells is assigned to the cell type with the highest assignment score to identify the tissue type origin of each cell. In step 218, the sequencing read counts are optionally trimmed to remove ambiguous calls. In step 220, the sequencing read count annotations are optionally smoothed using K-nearest neighbor smoothing.

[0022] In some embodiments, the single-cell RNA sequencing (scRNA-seq) data annotation pipeline disclosed herein is performed in three sequential steps: (1) scoring; (2) trimming; and (3) smoothing (Figures 3B-3D). Each step is described in further detail below. In some embodiments, the input to the methods disclosed herein consists of (1) a single-cell transcriptomics dataset (scRNA-seq or snRNA-seq data), (2) a set of cell types to be detected, and (3) a set of associated genes for each cell type. Methods for generating scRNA-seq data are reviewed in detail, for example, in Olsen TK and Baryawno N. Introduction to Single-Cell RNA Sequencing. Curr Protoc Mol Biol. 2018 Apr;122(1). The entire contents of this document are incorporated by reference. As shown in Figure 3A, the expression of single genes (rows) is assessed for each single cell (column) in the experiment based on the scRNA-seq data. Methods for inferring gene expression levels from scRNA-seq data are reviewed in detail, for example, in Birnbaum KD. Power in Numbers: Single-Cell RNA-Seq Strategies to Dissect Complex Tissues. Annu Rev Genet. 2018 Nov 23;52:203-221, which is incorporated herein by reference in its entirety. The method disclosed herein uses a score-based procedure to infer the cell type of origin of each cell based on a provided gene set. As shown in Figure 3B, single cells (columns) are scored against a provided gene set (rows). More specifically, for a given cell, the method first sorts genes with non-zero expression from high to low expression, generating a ranked vector. This ranked vector (length N) is then converted into a binary array (s) according to the following formula: where genes (g) included in a particular gene set (G) are replaced by 1 and all others by 0.

number

[0023] Next, a partial cumulative sum is performed over the binary array up to each element “k” of the binary vector, and then a sum is performed over all generated consecutive “1 to k” partial sums, resulting in an assignment score S for the given cell.

number

[0024] This process is performed across all cells and gene sets (M gene sets in total) to convert the input gene x cell expression matrix (number of genes x number of cells) into a gene set x cell assignment score matrix (number of gene sets x number of cells). Finally, each cell is assigned to the cell type with the highest assignment score. max{S m} 1≦m≦M·

[0025] trimming To prevent erroneous assignment when the cell type is unknown (not specified by the marker gene set), the methods disclosed herein identify cells that do not belong to any of the provided cell types (Figure 3C). In some embodiments, the methods disclosed herein utilize the Gini index as a measure of dispersion to identify ambiguous calls. According to the methods disclosed herein, first, for each cell, the Gini index (see, e.g., Jiang L, et al. GiniClust: detecting rare cell types from single-cell gene expression data with Gini index. Genome Biol. 2016 Jul 1;17(1):144., which is incorporated herein by reference in its entirety) is calculated between its assignment scores, and the gene set x cell assignment score matrix is ​​transformed into a Gini index (∈[0,1]) distribution. Next, cells that are outliers and have an index less than 0.5 are determined to be ambiguous. Such ambiguous cells are left unclassified. Additionally, cells that do not express a specified marker or that receive the same score in multiple cell types will similarly remain unclassified.

[0026] smoothing Next, k-nearest neighbor (kNN) smoothing (Figure 3D) is performed. This step is designed to smooth the annotations by aggregating information from cells (neighbors) with similar genome-wide expression profiles. Smoothing is performed by first identifying cells with a minimum of k nearest neighbors. Then, if the labels of more than 50% of the neighbors reach a consensus, the surrounded cell is relabeled according to the consensus of the neighbors. Smoothing is performed only once for all cells.

[0027] Other features By default, signature genes are expected to be highly expressed in one cell type compared to all other cell types. However, depending on the underlying data, these reference markers alone may not be sufficient to separate cell types with similar expression profiles. In such cases, genes that are expected to be undetectable in a particular cell type (e.g., CD8A in CD4 T cells) can be used to improve separation. Therefore, genes known to be characteristically less expressed in one cell type than another are introduced as "negative markers." In some embodiments, the methods disclosed herein incorporate negative markers by rewarding cells that do not express these markers and penalizing cells that do not. This series of steps can increase the variance between assignment scores and improve the accuracy of the method.

[0028] Sub-annotation strategies can be applied to approaches based on a hierarchy of known cell types. In some embodiments, data undergo a first round of annotation at intermediate hierarchical levels (i.e., broad cell type definitions). Subsequently, in some embodiments, each annotated group can be treated as a new, separate dataset and further annotated with additional gene sets. In some embodiments, applying the methods disclosed herein as a successive hierarchical process increases the granularity and precision of the method, allowing for more efficient discovery of novel cell subtypes.

[0029] Tissues and cell types in which the methods disclosed herein can be used to identify the cell type of single cells based on scRNA-seq data include peripheral blood mononuclear cells (PBMCs), including T cells and NK cells (TNKs), monocytes and macrophages (MPhs), B cells and plasma cells (BPCs), neutrophils (Neuts), erythrocytes (Eryths), megakaryocytes (Mega), and hematopoietic stem cells (HSCs); macrophages and monocytes, including classical (CD14+, CD16-), intermediate (CD14+, CD16+), and non-classical (CD14-, CD16+) monocytes; B cells, including B cells (including naive B cells), memory C cells (MBCs), and plasma cells (PCs); cardiac fibroblasts (CFs), cardiomyocytes (CMCs), endothelial cells (Endo), hepatocytes (Hepa), macrophages (Mac), and smooth muscle cells. cardiac tissue, kidney tissue, and lung tissue containing alveolar cells (SMC); kidney tissue containing B cells (B), endothelial cells (Endo), epithelial cells (Epit), macrophages (Mac), T and natural killer cell populations (TNK); and lung tissue containing adventitial cells (AC), alveolar type 1 and type 2 cells (ATC), basal cells (Basal), basophils (Baso), B and plasma cells (BPC), club cells (CC), endothelial cells (Endo), fibroblasts (Fibro), goblet cells, serous cells, mucus cells (GSM), pulmonary ciliated cells (LCC), mesothelial cells (MC), monocytes, dendritic cells, macrophages (MPh), neutrophils (Neut), pericytes (PC), lung ionocytes (PI), smooth muscle cells (SMC), T cells, and natural killer cells (TNK).

[0030] In some embodiments, tissues in which the methods disclosed herein can be used to identify the cell type of a single cell based on scRNA-seq data include, but are not limited to, prostate, lung, pancreas, cervix, kidney, salivary gland, uterus, stomach, thyroid, paranasal sinuses, middle and inner ear, adrenal gland, appendix, hematopoietic system, bones and joints, spinal cord, breast, cerebellum, connective and soft tissue, uterine body, esophagus, eye, nose, eye, fallopian tube, extrahepatic bile duct, mouth, intrahepatic bile duct, kidney, appendix, larynx, lips, liver, lung and bronchi, lymph nodes, cerebrum, spinal cord, nasal cartilage, retina, oropharynx, endocrine glands, female reproductive organs, ovaries, penis and scrotum, pituitary gland, pleura, rectum, renal pelvis, ureter, peritoneum, salivary gland, skin, small intestine, testis, thymus, thyroid, tongue, unknown, bladder, uterus, vagina, labia, vulva tissue. In some embodiments, the sample comprises cells selected from the group consisting of adipose, adrenal cortex, adrenal gland, adrenal medulla, appendix, bladder, blood, blood vessel, bone, osteochondral, brain, breast, cartilage, cervix, colon, colon-sigmoid, dendritic cell, skeletal muscle, endometrium, esophagus, fallopian tube, fibroblast, gallbladder, kidney, larynx, liver, lung, lymph node, melanocyte, mesothelial lining, myoepithelial cell, osteoblast, ovary, pancreas, parotid gland, prostate, salivary gland, sinus tissue, skeletal muscle, skin, small intestine, smooth muscle, stomach, synovium, endoarticular tissue, tendon, testis, thymus, thyroid, uterus, and uterine corpus. In some embodiments, the tissue is obtained from a healthy subject. In some embodiments, the tissue is obtained from a subject known or suspected to have cancer. In some embodiments, the tissue is obtained from a solid or liquid tumor.

[0031] The list of cell types and associated markers for specifying gene sets can be obtained from a variety of sources. In some embodiments, the methods disclosed herein include providing a user-specified list of gene sets. In some embodiments, the cell types and associated markers expected to be observed in the tissue under investigation are provided in advance. Sources of lists of tissue-specific markers include, for example, CellMarker (Zhang X, et al. CellMarker: a manually curated resource of cell markers in human and mouse. Nucleic Acids Res. 2019 Jan 8;47(D1):D721-D728. doi:10.1093 / nar / gky900. PMID:30289549; PMCID:PMC6323899., which is incorporated herein by reference in its entirety) and PanglaoDB (Franzen O, et al. PanglaoDB: a web server for exploration of mouse and human single-cell RNA sequencing data. Database (Oxford). 2019 Jan 1;2019., which is incorporated herein by reference in its entirety). Expert-annotated data is also available from databases including, for example, The Tabula Sapiens (Tabula Sapiens Consortium. The Tabula Sapiens: A multiple-organ, single-cell transcriptomic atlas of humans. Science. 2022 May 13;376(6594), incorporated herein by reference in its entirety), and in some embodiments can be analyzed using the methods disclosed herein to define gene signatures.Signatures can be defined by the user's own experiments, such as CITE-seq (Stoeckius M, et al. Simultaneous epitope and transcriptome measurement in single cells. Nat Methods. 2017 Sep;14(9):865-868, which is incorporated herein by reference in its entirety) or cross-validation experiments.

[0032] The methods disclosed herein can be used as standalone tools or as a complement to other supervised methods. For example, the methods disclosed herein can be used in combination with scRNA-seq analysis workflows, such as the Seurat package (Hao Y, et al. Integrated analysis of multimodal single-cell data. Cell. 2021 Jun 24;184(13):3573-3587., which is incorporated herein by reference in its entirety), to identify separated cell populations, after which the methods disclosed herein can be applied to further refine the granularity of cell type annotation.

[0033] Computer implementation of the method FIG. 5 is a component diagram of a computer system 500 that can be used to implement a method for identifying the cell type origin of each cell of a plurality of cells based on scRNA-seq data.

[0034] Computing device 500 is intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other suitable computers. Computing device 550 is intended to represent various forms of mobile devices, such as personal digital assistants, mobile phones, smartphones, and other similar computing devices. Additionally, computing device 500 or 550 may include a universal serial bus (USB) flash drive. The USB flash drive may store an operating system and other applications. The USB flash drive may include input / output components, such as a wireless transmitter or a USB connector that can be inserted into a USB port of another computing device. The components, their connections and relationships, and their functions illustrated herein are intended to be exemplary only and are not intended to limit the implementation of the methods and configurations described and / or claimed herein.

[0035] The computing device 500 includes a processor 502, memory 504, a storage device 506, a high-speed interface 508 connecting to the memory 504 and a high-speed expansion port 510, a low-speed bus 514, and a low-speed interface 512 connecting to the storage device 506. The components 502, 504, 506, 508, 510, and 512 are interconnected using various buses and may be implemented on a common motherboard or in other suitable manners. The processor 502 processes instructions for execution within the computing device 500, including instructions stored in the memory 504 or the storage device 506, and can display graphical information for a GUI on an external input / output device, such as a display 516 coupled to the high-speed interface 508. Other implementations may use multiple processors and / or multiple buses, along with multiple memories and types of memory, as appropriate. Multiple computing devices 500 may also be connected, each providing a portion of the necessary operations, for example, as a server bank, a cluster of blade servers, or a multiprocessor system.

[0036] The memory 504 stores information within the computing device 500. In one implementation, the memory 504 is one or more volatile memory units. In another embodiment, the memory 504 is one or more non-volatile memory units. The memory 504 may also be another form of computer-readable medium, such as a magnetic disk or optical disk.

[0037] The storage device 506 can provide mass storage for the computing device 500. In one embodiment, the storage device 506 can be or include a computer-readable medium such as a floppy disk device, a hard disk device, an optical disk device, or a tape device, a flash memory or other similar solid-state memory device, or an array of devices including devices in a storage area network or other configuration. A computer program product can be tangibly embodied on an information carrier. The computer program product can also include instructions that, when executed, perform one or more methods, such as those described above. The information carrier is a computer- or machine-readable medium, such as memory 504, the storage device 506, or memory on the processor 502.

[0038] The high-speed controller 508 manages bandwidth-intensive operations of the computing device 500, while the low-speed controller 512 manages less bandwidth-intensive operations. This division of functions is merely an example. In one embodiment, the high-speed controller 508 is coupled to the memory 504, the display 516, via a graphics processor or accelerator, for example, and to a high-speed expansion port 510 that can accept various expansion cards (not shown). In this embodiment, the low-speed controller 512 is coupled to the storage device 506 and the low-speed expansion port 514. The low-speed expansion port may include various communication ports, such as USB, Bluetooth, Ethernet, wireless Ethernet, etc., but may also be coupled via a network adapter to one or more input / output devices, such as a keyboard, a pointing device, a microphone / speaker pair, a scanner, or a networking device, such as a switch or router. The computing device 500, as shown, can be implemented in several different forms. For example, it can be implemented as a standard server 520 or multiple times within a cluster of such servers. It can also be implemented as part of a rack server system 524. Additionally, it may be implemented in a personal computer such as a laptop computer 522. Alternatively, components from computing device 500 may be combined with other components in a mobile device (not shown), such as device 550. Each such device may include one or more computing devices 500, 550, and the overall system may be made up of multiple computing devices 500, 550 in communication with each other.

[0039] Computing device 500, as shown, can be implemented in several different forms. For example, it can be implemented as a standard server 520 or multiple times within a cluster of such servers. It can also be implemented as part of a rack server system 524. Furthermore, it can be implemented in a personal computer, such as a laptop computer 522. Alternatively, components from computing device 500 can be combined with other components in a mobile device (not shown), such as device 550. Each such device can include one or more computing devices 500, 550, and the entire system can be made up of multiple computing devices 500, 550 communicating with each other.

[0040] Computing device 550 includes, among other components, a processor 552, memory 564, input / output devices such as a display 554, a communication interface 566, and a transceiver 568. Device 550 may be provided with a storage device such as a microdrive or other device to provide additional storage. Each of the components 550, 552, 564, 554, 566, 568 are interconnected using various buses, and some of the components may be implemented on a common motherboard or in other suitable manner.

[0041] The processor 552 can execute instructions within the computing device 550, including instructions stored in memory 564. The processor can be implemented as a chipset of chips including separate analog and digital processors. Additionally, the processor can be implemented using any of a number of architectures. For example, the processor 510 can be a CISC (Complex Instruction Set Computer) processor, a RISC (Reduced Instruction Set Computer) processor, or a MISC (Minimum Instruction Set Computer) processor. The processor can coordinate other components of the device 550, such as controlling a user interface, applications run by the device 550, and wireless communication by the device 550.

[0042] The processor 552 can communicate with a user via a control interface 558 and a display interface 556 coupled to a display 554. The display 554 can be, for example, a TFT (thin film transistor liquid crystal display) display, an OLED (organic light emitting diode) display, or other suitable display technology. The display interface 556 can include appropriate circuitry for driving the display 554 to present graphical and other information to the user. The control interface 558 can receive commands from the user and translate them for submission to the processor 552. Additionally, an external interface 562 can be in communication with the processor 552 to enable short-range communication between the device 550 and other devices. The external interface 562 can provide, for example, for wired communication in some implementations or wireless communication in other implementations; multiple interfaces can also be used.

[0043] Memory 564 stores information within computing device 550. Memory 564 may be implemented as one or more computer-readable media, one or more volatile memory units, or one or more non-volatile memory units. Expansion memory 574 may also be provided and connected to device 550 via expansion interface 572, which may include, for example, a SIMM (single in-line memory module) card interface. Such expansion memory 574 may provide additional storage space for device 550 or may store applications or other information for device 550. Specifically, expansion memory 574 may include instructions for implementing or supplementing the processes described above, and may also include secure information. Thus, for example, expansion memory 574 may be provided as a security module for device 550 and programmed with instructions that enable secure use of device 550. Additionally, secure applications may be provided via SIMM cards, along with additional information, such as placing identifying information on the SIMM card in an unhackable manner.

[0044] The memory may include, for example, flash memory and / or NVRAM memory, as described below. In one embodiment, a computer program product is tangibly embodied on an information carrier. The computer program product includes instructions that, when executed, perform one or more methods, such as those described above. The information carrier is, for example, a computer-readable or machine-readable medium, such as memory 564, expansion memory 574, or memory on processor 552, receivable via transceiver 568 or external interface 562.

[0045] Device 550 can communicate wirelessly via communication interface 566, which may include digital signal processing circuitry as needed. Communication interface 566 can provide for communication in a variety of modes or protocols, such as GSM voice calls, SMS, EMS, or MMS messaging, CDMA, TDMA, PDC, WCDMA, CDMA2000, or GPRS. Such communication can occur, for example, via radio frequency transceiver 568. Additionally, short-range communication can occur, such as using Bluetooth, Wi-Fi, or other such transceivers (not shown). Additionally, GPS (Global Positioning System) receiver module 570 can provide additional navigation- and location-related wireless data to device 550, which can be used as appropriate by applications executing on device 550.

[0046] Device 550 can also communicate voice using audio codec 560, which can receive voice information from a user and convert it into usable digital information. Audio codec 560 can also generate audible sounds for the user, for example, through a speaker in the handset of device 550. Such sounds can include sounds from a voice call, recorded sounds such as voice messages, music files, and the like, and can also include sounds generated by applications running on device 550.

[0047] The computing device 550, as shown, can be implemented in several different forms, for example, as a mobile phone 580. It can also be implemented as part of a smartphone 582, personal digital assistant, or other similar mobile device.

[0048] Various implementations of the systems and methods described herein may be realized in digital electronic circuitry, integrated circuits, specially designed ASICs (application-specific integrated circuits), computer hardware, firmware, software, and / or combinations of such implementations. These various implementations may include implementation in one or more computer programs executable and / or interpretable by a programmable system including at least one programmable processor, which may be special purpose or general purpose, coupled to receive data and instructions from, and transmit data and instructions to, a storage system, at least one input device, and at least one output device.

[0049] These computer programs (also referred to as programs, software, software applications, or code) include machine instructions for a programmable processor and may be implemented in a high-level procedural and / or object-oriented programming language and / or assembly / machine language. As used herein, the terms "machine-readable medium" and "computer-readable medium" refer to any computer program product, apparatus, and / or device used to provide machine instructions and / or data to a programmable processor, including a machine-readable medium that receives machine instructions as a machine-readable signal, such as a magnetic disk, optical disk, memory, or programmable logic device (PLD). The term "machine-readable signal" refers to any signal used to provide machine instructions and / or data to a programmable processor.

[0050] To provide for user interaction, the systems and techniques described herein can be implemented in a computer having a display device, such as a CRT (cathode ray tube) or LCD (liquid crystal display) monitor, for displaying information to a user, and a keyboard and pointing device, such as a mouse or trackball, by which the user can provide input to the computer. Other types of devices can be used to provide for user interaction as well; for example, feedback provided to the user can be any form of sensory feedback, such as visual feedback, auditory feedback, or tactile feedback, and input from the user can be received in any form, including acoustic, speech, or tactile input.

[0051] The systems and techniques described herein may be implemented in a computing system that includes back-end components, such as, for example, a data server, or includes middleware components, such as an application server, or includes front-end components, such as a client computer having a graphical user interface or web browser that allows a user to interact with an implementation of the systems and techniques described herein, or any combination of such back-end, middleware, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication, such as a communications network. Examples of communications networks include a local area network ("LAN"), a wide area network ("WAN"), and the Internet.

[0052] A computing system may include clients and servers. Clients and servers are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other.

[0053] A number of embodiments have been described. Nevertheless, it will be understood that various modifications are possible without departing from the spirit and scope of the present invention. Additionally, the logic flows depicted in the figures do not require the particular order shown, or sequential order, to achieve desirable results. Additionally, other steps may be provided from or steps may be deleted from the described flows, and other components may be added to or removed from the described systems. Accordingly, other embodiments are within the scope of the following claims.

[0054] All of the embodiments and functional operations of the present disclosure described herein may be implemented in digital electronic circuitry, or computer software, firmware, or hardware, including the structures disclosed herein and their structural equivalents, or in one or more combinations thereof. Embodiments of the methods and compositions may be implemented as one or more computer program products, e.g., one or more modules of computer program instructions encoded on a computer-readable medium for execution by or to control the operation of a data processing apparatus. The computer-readable medium may be a machine-readable storage device, a machine-readable storage substrate, a memory device, a composition of matter providing a machine-readable propagated signal, or one or more combinations thereof. The term "data processing apparatus" encompasses all apparatuses, devices, and machines for processing data, including, by way of example, a programmable processor, a computer, or multiple processors or computers. In addition to hardware, an apparatus may include code that creates an execution environment for the computer program in question, such as code constituting processor firmware, a protocol stack, a database management system, an operating system, or one or more combinations thereof. A propagated signal is an artificially generated signal, e.g., a machine-generated electrical, optical, or electromagnetic signal generated to encode information for transmission to an appropriate receiver apparatus.

[0055] A computer program (also known as a program, software, software application, script, or code) may be written in any form of programming language, including compiled or interpreted languages, and may be deployed in any form, such as a stand-alone program or as a module, component, subroutine, or other unit suitable for use in a computing environment. A computer program does not necessarily correspond to a file in a file system. A program may be stored as part of a file that holds other programs or data (e.g., one or more scripts stored in a markup language document), in a single file dedicated to the program in question, or in multiple associated files (e.g., files storing one or more modules, subprograms, or portions of code). A computer program may be deployed to be executed on one computer or on multiple computers that are located at one site or distributed across multiple sites and interconnected by a communications network.

[0056] The processes and logic flows described herein may be performed by one or more programmable processors that execute one or more computer programs to perform functions by operating on input data and generating output. The processes and logic flows may also be performed by, and apparatus may be implemented as, special purpose logic circuitry, such as an FPGA (Field Programmable Gate Array) or an ASIC (Application Specific Integrated Circuit).

[0057] Processors suitable for executing a computer program include, by way of example, both general-purpose and special-purpose microprocessors, as well as any one or more processors of any kind of digital computer. Generally, a processor receives instructions and data from a read-only memory or a random-access memory, or both. The essential elements of a computer are a processor for executing instructions and one or more memory devices for storing instructions and data. Generally, a computer will include one or more mass storage devices, such as magnetic, magneto-optical, or optical disks, or both, for storing data, or will be operatively coupled to receive or transfer data from such devices. However, a computer need not have such devices. Furthermore, a computer can be incorporated into another device, such as a tablet computer, a mobile phone, a personal digital assistant (PDA), a portable audio player, or a global positioning system (GPS) receiver, to name a few. Computer-readable media suitable for storing computer program instructions and data include all forms of non-volatile memory, media, and memory devices, including, by way of example, semiconductor memory devices such as EPROM, EEPROM, and flash memory devices; magnetic disks, such as internal hard disks or removable disks; magneto-optical disks; and CD-ROM and DVD-ROM disks. The processor and the memory can be supplemented by, or incorporated in, special purpose logic circuitry.

[0058] To provide for user interaction, embodiments of the present disclosure can be implemented on a computer having a display device, such as a CRT (cathode ray tube) or LCD (liquid crystal display) monitor, for displaying information to the user, and a keyboard and pointing device, such as a mouse or trackball, by which the user can provide input to the computer. Other types of devices can be used to provide for user interaction as well; for example, feedback provided to the user can be any form of sensory feedback, such as visual feedback, auditory feedback, or tactile feedback, and input from the user can be received in any form, including acoustic, speech, or tactile input.

[0059] Embodiments of the present disclosure may be implemented in a computing system including a back-end component, e.g., a data server or middleware component, a computing system including an application server, or a front-end component, e.g., a computing system including a client computer having a graphical user interface or web browser through which a user can interact with method embodiments, or any combination of one or more such back-end, middleware, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication, e.g., a communications network. Examples of communications networks include a local area network ("LAN"), a wide area network ("WAN") such as the Internet, or the like.

[0060] A computing system may include clients and servers. Clients and servers are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other.

[0061] While this specification contains many details, these should not be construed as limitations on the scope of the invention or what may be claimed, but rather as descriptions of features specific to particular embodiments of the invention. Certain features described herein in the context of separate embodiments may also be implemented in combination in a single embodiment. Conversely, various features described in the context of a single embodiment may also be implemented in multiple embodiments separately or in any suitable subcombination. Furthermore, while features may be described above as acting in a particular combination, and even initially claimed as such, one or more features from a claimed combination may, in some cases, be carved out of that combination, and the claimed combination may be directed to a subcombination or a variation of the subcombination.

[0062] Similarly, while operations are shown in the figures in a particular order, this should not be understood as requiring that such operations be performed in the particular order shown, or in any sequential order, or that all of the illustrated operations be performed, to achieve desirable results. In certain situations, multitasking and parallel processing may be advantageous. Furthermore, the separation of various system components in the above-described embodiments should not be understood as requiring such separation in all embodiments, and it should be understood that the described program components and systems may generally be integrated into a single software product or packaged into multiple software products.

[0063] In each instance where an HTML file is mentioned, other file types or formats can be substituted. For example, the HTML file can be replaced with an XML, JSON, plain text, or other type of file. Additionally, where a table or hash table is mentioned, other data structures (such as a spreadsheet, relational database, or structured file) can be used. [Example]

[0064] The compositions and methods are further described in the following examples, which do not limit the scope of the claimed compositions and methods.

[0065] Example 1: Cell type identification using simulated data The performance of the cell-type identification method was benchmarked using simulated data sets in which cell types were known in advance. Using the R package Splatter (Zappia L, et al., Splatter: Simulation of Single-Cell RNA Sequencing Data. Genome Biol. 2017 Sep 12;18(1):174.) with default parameters, we generated three simulated scRNA-seq datasets of increasing size and complexity: (1) six datasets with 5,000 cells divided into five cell types, (2) six datasets with 10,000 cells divided into 10 cell types, and (3) six datasets with 15,000 cells divided into 15 cell types. The proportion of cells in each cell type group was randomly sampled from a uniform distribution. Each simulated dataset contained 10,000 genes, as shown in Figure 4A.

[0066] For benchmarking using simulated data, cell type identification methods used cell type-specific markers as input. A systematic approach was designed to infer cell type-specific markers from simulated data. First, one simulated data set was selected as the training data set (1 training + 5 testing). Using the FindAllMarkers function in the Seurat R package (Hao Y, et al. Integrated analysis of multimodal single-cell data. Cell. 2021 Jun 24;184(13):3573-3587.e29.), differential expression analysis using the Wilcoxon rank-sum test was performed to infer marker gene sets from the training data if the proportion of cells expressing a given gene was 0.1 or greater and there was at least a 0.5-fold difference (logarithmic scale) between the two cell groups. The top 100 marker genes for each cell type were used as input to annotate the remaining five testing data sets. This approach provided a benchmark of 30 trials for each set of simulation data, as shown in Figure 4B.

[0067] As shown in Figure 4C, the performance of the cell type identification method using simulated data is quantified using measures of sensitivity (SEN), specificity (SPC) and F1 score, expressed as fractions of TP / (TP+FN), TN / (TN+FP) and 2×TP / (2×TP+FP+FN), respectively: 1. True positives (TP) were defined as the number of correctly identified cell-type related pairs. 2. False positives (FP) were defined by the number of unrelated cell pairs incorrectly identified as cell-type related. 3. True negatives (Tn) were defined by the number of unrelated cell pairs correctly identified as unrelated. 4. False negatives (FN) were defined by the number of cell type-associated pairs incorrectly identified as unrelated.

[0068] As shown in the plots in Figure 4C, the cell type identification method inferred simulated cell type assignments with an average sensitivity, specificity, and F1 score of over 99% across all trials.

[0069] The performance of the cell type identification method was evaluated for the discovery of novel cell types (i.e., unclassified cell types). This was done by omitting a subset of the marker gene set and applying the cell type identification method to extract the missing cell types. One training dataset was randomly selected, and 1–4 cell types were randomly selected from the training data: 1–2 cell types from the 5-cell type data, 1–3 cell types from the 10-cell type data, and 1–4 cell types from the 15-cell type data. Next, marker gene sets were inferred from the training data by performing differential expression analysis as described above. Finally, the remaining simulation data was used to benchmark the cell type identification method. This procedure was repeated 100 times for each cell type removal. Cell types were removed before selecting marker genes, ensuring that marker genes were selected without knowing the unknown cell types. As shown in Figure 4D, the cell type identification method predicted unknown cell types with an average sensitivity, specificity, and F1 score of over 95% across all tests.

[0070] The cell type identification method is not affected by the inclusion of marker gene sets for cell types not present in the dataset. Because the cell type identification method is a single-cell-based algorithm, the inclusion of a marker set for cells not found does not affect the scores of other marker gene sets and therefore does not affect the quality of annotation of cell types present in the dataset.

[0071] Example 2: Cell type identification using experimental data The performance of the cell type identification method disclosed herein was evaluated by annotating experimental scRNA-seq data from multiple human tissues published by the Tabula Sapiens Consortium (Tabula Sapiens Consortium. The Tabula Sapiens: A multiple-organ, single-cell transcriptomic atlas of humans. Science. 2022 May 13;376(6594):eabl4896.). 50,115 peripheral blood mononuclear cells (PBMCs) were extracted from the Tabula data and classified using a hierarchy of known immune cell types. First, we created the following marker gene sets that identify cell types with moderate immunophenotypic granularity: T cells and NK cells (TNK), monocytes and macrophages (MPh), B cells and plasma cells (BPC), neutrophils (Neut), erythrocytes (Eryth), megakaryocytes (Mega), and hematopoietic stem cells (HSC). Marker genes for T cells and NK cells included CD4, CD3D, CD3E, CD3G, IL7R, CD8A, CD8B, FOXP3, TIGIT, CD27, NCAM1, KLRF1, GNLY, NKG7, and TNFRSF18. Marker genes for monocytes and macrophages included CD4, MS4A7, VCAN, CD14, FCN1, LYZ, FCGR3A, ITGAM, MARCO, CD1C, and ITGAX. Marker genes for B cells and plasma cells included CD79A, CD79B, MS4A1, CD19, CD24, CD37, CD72, TNFRSF17, MZB1, CR2, BCL11A, FCRL1, FCRL2, FCRL5, SPIB, FCRLA, and JCHAIN. Neutrophil marker genes included S100A8, S100A9, IFITM2, FCGR3B, MS4A2, CPA3, TPSAB1, FCER1A, and MPO. Erythroid marker genes included HBB, HBA1, and HBA2. Megakaryocyte marker genes included PPBP, PF4, and ITGA2B. Hematopoietic stem cell marker genes included CD34, PROM1, SPINK2, HOPX, CLEC9A, and SOX4.

[0072] As shown in Figure 6A, the cell-typing method assigned cells to the correct type (expert-annotated type) with 98% accuracy for all cell types. As shown in Figure 6B, the Jaccard index (JI) also demonstrated high assignment accuracy: 98% for MPhs, 96% for TNK, BPC, and Eryts, 94% for Neuts, and 76% for Megas. A relatively low JI was achieved for HSCs: 39%. As shown in Figure 6C, the expression levels of the HSC reference markers CD34, CD133, and SPINK2 indicated that cells annotated by the cell-typing method disclosed herein were more likely to be HSCs than cells annotated manually. Only 16 of the more than 50,000 cells did not express all of the immunophenotypic reference markers and therefore could not be assigned to a cell type by the cell-typing method.

[0073] The performance of the cell-typing method was evaluated by increasing the granularity of annotated cell types within the PBMC cell population. Using the cell-typing method disclosed herein and the Tabula test, 10,173 cells co-annotated as T cells or natural killer cells (TNK) were retrieved from the Tabula Consortium data. Next, a list of reference markers for T cells and NK (natural killer) cells was created. The reference markers are shown in Figure 7A. T cell marker genes included CD4, CD3D, CD3E, CD3G, IL7R, CD8A, CD8B, FOXP3, TIGIT, CD27, NKG7, and GNLY. Natural killer cell marker genes included KLRF1, KLRC1, KLRC2, NCAM1, FGFBP2, FCGR3A, CX3CR1, NCR1, FCER1G, KDELC1, NKG7, GNLY, KLRD1, and CD7.

[0074] As shown in Figure 7B, the cell-type identification method clearly annotated 5,683 T cells and 4,486 NK cells in t-SNE space, leaving only four cells unannotated (Figure 7C). As shown in Figure 7D, the Jaccard index similarity between the cell-type identification method and the Tabula test was 75% for T cells and 59% for NK cells. A mixed population (18% JI) of 1,795 cells (17% of TNK) was identified as NK cells by the cell-type identification method disclosed herein and as T cells by the Tabula test (Figure 7D). This mixed-annotation population was investigated to determine which of the two annotations was more appropriate. The expression levels of T cell markers, including CD3D, CD4, and CD8A, were examined. As shown in Figure 7E, this mixed population did not express any T cell markers. In contrast, they highly expressed cytotoxic markers such as NKG7, GNLY, FGFBP2, FCGR3A, and FCER1G. Therefore, the NK cell annotation predicted by the cell type identification method disclosed herein was more reasonable than the T cell annotation in the Tabula test.

[0075] Next, we improved the resolution of the cell-typing method by focusing only on T cells. 5,634 cells co-annotated with T cells using the cell-typing method disclosed herein and the Tabula test were retrieved. We created a list of reference markers for CD4+ and CD8+ T cells (Figure 8A). Marker genes for CD4+ T cells included CD3D, IL7R, and CD4, while negative marker genes for CD4+ T cells included CD8A and CD8B. Marker genes for CD8+ T cells included CD3D, IL7R, CD8A, and CD8B, while negative marker genes for CD8+ T cells included CD4.

[0076] As shown in Figures 8B and 8C, the cell-typing method annotated 3,995 CD4+ T cells and 1,639 CD8+ T cells. As shown in Figure 8D, the Jaccard index similarity between the cell-typing method and the Tabula test was 78% for CD4+ T cells and 51% for CD8+ T cells. In addition, a mixed population (17% of TCs) was observed among the cells annotated by the cell-typing method and the Tabula test. This consisted of 376 cells (7% JI) annotated as CD4+ T cells by the cell-typing method but CD8+ T cells by the Tabula test, and vice versa (605 cells (12% JI)) (Figure 8D). These two mixed populations were investigated to determine the more appropriate annotation. The expression levels of the reference T cell markers CD4, CD8A, and CD8B were evaluated. As shown in Figure 8E, gene expression of these markers indicated that cells annotated as CD4+ T cells by the cell-typing method expressed the CD4 marker but not CD8A or CD8B (left panel). In contrast, cells annotated as CD8+ T cells by the Tabula test did not express the CD8A or CD8B markers. Furthermore, as shown in Figure 8E, cells annotated as CD8+ T cells by the cell-typing method expressed the CD8A and CD8B markers but not CD4 (right panel), whereas cells annotated as CD4+ T cells by the Tabula test expressed relatively high levels of the CD8A and CD8B markers but not CD4 marker. These results indicate that the annotation by the cell-typing method disclosed herein is more accurate than the Tabula test for CD4+ and CD8+ subpopulations.

[0077] Next, we evaluated the annotation of macrophages and monocytes using the cell-typing method disclosed herein. The performance of the cell-typing method was benchmarked by annotating macrophages and three major monocyte populations: classical (CD14+, CD16-), intermediate (CD14+, CD16+), and atypical (CD14-, CD16+) monocytes. 16,504 cells co-annotated with MPh were retrieved using both the cell-typing method and the Tabula test. A list of reference markers for macrophages and monocytes was created. Macrophage marker genes included ITGAM, ITGAX, CD68, MRC1, FCGR1A, and FCGR2A, while negative macrophage marker genes included CD14, FCGR3A, CDKN1C, RHOC, LYZ, and MS4A7. Marker genes for classical (CD14+, CD16-) monocytes include CD14 and LYZ, and negative marker genes for classical monocytes include FCGR3A, MS4A7, CD68, CDKN1C, and RHOC. Marker genes for intermediate (CD14+, CD16+) monocytes include CD14 and FCGR3A, and negative marker genes for intermediate monocytes include CD68. Marker genes for non-classical (CD14-, CD16+) monocytes include FCGR3A and MS4A7, and negative marker genes for non-classical monocytes include CD14, LYZ, and CD68.

[0078] As shown in Figures 9A and 9B, the cell-typing method identified 894 macrophages, 10,837 classical monocytes, 4,719 intermediate monocytes, 96 non-classical monocytes, and 3 unlabeled monocytes. As shown in Figure 9C, the similarity between the annotations from the cell-typing method disclosed herein and those from the Tabula test, measured by the Jaccard index, indicated low agreement. As with T cells and NK cells above, the accuracy of each annotation was evaluated. First, the expression levels of macrophage reference markers, such as ITGAM, ITGAX, CD68, FCGR1A, and FCGR2A, were evaluated. This evaluation indicated that macrophages annotated by the cell-typing method disclosed herein expressed higher levels of these reference markers than macrophages annotated by the Tabula test (Figure 9B). Further examination of the monocyte population, as shown in Figure 9D, demonstrated that the cell-type identification method disclosed herein achieved a clear separation between classical, intermediate, and non-classical monocytes (top panel). In contrast, cells annotated by the Tabula test displayed a mixed population, indicating a failed annotation (bottom panel). Thus, these data indicate that cells annotated by the cell-type identification method disclosed herein are more likely to be macrophages or monocytes than cells annotated manually.

[0079] Next, we benchmarked the performance of the cell-typing method disclosed herein on B cell populations, including naive B cells (B naive), memory B cells (MBC), and plasma cells (PC). 3,525 cells co-annotated with BPCs by the cell-typing method and Tabula were searched. We created a list of reference markers highlighting memory, naive, and plasma states. Marker genes for naive B cells included CD72, CD69, CD24, IGHM, and IGHD, while negative marker genes for naive B cells included MZB1, PRDM1, XBP1, IGHG1, IGHG2, IGHG3, IGHA1, CD27, and MME. Marker genes for memory B cells included CD72, CD69, CD24, IGHG1, IGHG2, IGHG3, IGHA1, IGHA2, IGHE, MME, and CD27, and negative marker genes for memory B cells included MZB1, PRDM1, XBP1, IGHM, and IGHD.Marker genes for plasma cells included PRDM1, XBP1, JCHAIN, and MZB1, and negative marker genes for plasma cells included CD72, CD69, and CD24.

[0080] As shown in Figures 10A and 10B, the cell-type identification method annotated 2,486 NBCs, 593 MBCs, and 446 PCs. As shown in Figure 10C, the Jaccard index similarity between the cell-type identification method and the Tabula test annotations was 87% for NBCs, 64% for MBCs, and 99% for PCs. A mixed population (9% JI) consisting of 284 cells (8% of BPCs) was identified as consisting of cells annotated as NBCs by the cell-type identification method and MBCs by the Tabula test (Figure 10C). We examined the expression levels of reference markers for BCs, including the naive markers IGHM and IGHD and the activation marker CD27. As shown in Figure 10D, this population highly expressed IGHM and IGHD and did not express CD27. Therefore, these data indicate that the cell-type identification method's annotation of NBCs in this population is more likely to be accurate than the Tabula annotation.

[0081] Next, we evaluated the performance of our cell type identification method in assigning cell types for three additional human tissues (heart, kidney, and lung) by comparing it with manual cell type annotations using the Tabula test. The manual annotation for human heart tissue included six cell types: cardiac fibroblasts (CFs), cardiomyocytes (CMCs), endothelial cells (Endo), hepatocytes (Hepa), macrophages (Mac), and smooth muscle cells (SMCs). A set of marker genes was selected from the literature to recognize these cell types in the dataset. Marker genes for cardiac fibroblasts included DCN, GSN, PDGFRA, COL1A1, C7, LUM, and TCF21. Marker genes for cardiomyocytes included NPPA, MYL7, MYL4, TNNT2, and MYH6. Marker genes for endothelial cells included VWF, PECAM1, CDH5, FABP4, AQP7, NPR3, PLVAP, and KDR. Hepatocyte marker genes included TAGLN, AFP, KRT8, KRT18, HNF4A, LRP5, FGFR4, ASGR1, UCP2, HHEX, FOXA1, FOXA2, FOXA3, CDH1, FGA, FGB, FGG, APOA2, APOC3, and AMBP. Macrophage marker genes included CD163, C1QA, TAGLN, MARCO, MRC1, ITGAM, MS4A7, CD14, FCGR3A, CD4, S100A8, S100A9, C1QC, GPR183, LYZ, ITGAX, VSIG4, FCER1G, TLR2, CD68, PTPRC, CSF1R, ITGAX, VSIG4, CD86, DUSP1, and DUSP2. Smooth muscle cell marker genes included MYH11, TAGLN, ACTA2, CALD1, and NOTCH3.

[0082] As shown in Figures 11A and 11B, the cell-type identification method disclosed herein demonstrated 96% similarity to manually annotated cells across all six cardiac tissue cell types. As shown in Figure 11C, high assignment similarity between the cell-type identification method and the manual annotations of the Tabula test was demonstrated at the individual cell-type level as measured by the Jaccard index: 97% for CF, 96% for Endo cells, 95% for CMC, 85% for SMC, 70% for Hepa cells, and 67% for Mac.

[0083] Next, we evaluated the performance of the cell type identification method disclosed herein on human kidney tissue containing 9,461 cells. The original annotation included five cell types: B cells (B), endothelial cells (Endo), epithelial cells (Epit), macrophages (Mac), T cells, and natural killer cell populations (TNKs), and a set of marker genes was compiled for each. Marker genes for renal B cells included CD79A, CD79B, MS4A1, CD19, and JCHAIN. Marker genes for renal endothelial cells included TGFBR2, NOTCH4, ADGRL4, EMCN, ENG, PECAM1, PLVAP, TEK, KDR, EPCAM, GJA5, CDH5, SEMA3G, PTPRB, and SLC14A1. Renal epithelial cell marker genes included ANPEP, AQP1, SLC13A3, SLC16A9, SLC17A3, SLC22A7, and SLC22A8. Renal macrophage marker genes included CD4, MS4A7, VCAN, CD14, FCN1, LYZ, FCGR3A, ITGAM, MARCO, CD1C, and ITGAX. Renal T cell and natural killer cell marker genes included CD4, CD3D, CD3E, CD3G, IL7R, CD8A, CD8B, FOXP3, TIGIT, CD27, NCAM1, KLRF1, GNLY, NKG7, and TNFRSF18.

[0084] As shown in Figures 12A and 12B, the cell type identification method achieved a high similarity of 99% with the manual annotations of the Tabula test for all cell types, with only 17 unlabeled cells (Figure 12B). As shown in Figure 12C, the high assignment similarity between the cell type identification method and the manual annotations of the Tabula test was indicated by high Jaccard indices: 100% for Epit cells, 98% for TNK cells, 96% for B cells, 91% for Mac cells, and 74% for Endo cells.

[0085] Finally, we evaluated the performance of the cell type identification method described herein on complex human lung tissue (35,682 cells from the Tabula Sapiens study). Initial annotation included the following 17 cell types: adventitial cells (AC), alveolar type 1 and type 2 cells (ATC), basal cells (Bas), basophils (Baso), B cells and plasma cells (BPC), club cells (CC), endothelial cells (Endo), fibroblasts (Fibro), goblet cells, serous cells and mucus cells (GSM), pulmonary ciliated cells (LCC), mesothelial cells (MC), monocytes, dendritic cells, macrophages (MPh), neutrophils (Neut), pericytes (PC), lung ionocytes (PI), smooth muscle cells (SMC), T cells, and natural killer cells (TNK). A set of marker genes was collected from the literature to detect and annotate these cell types in the dataset. Marker genes for adventitial cells included CD34, ANPEP, UAP1, VIT, MFAP5, PCOLCE2, GFPT2, DPT, PDGFRA, LUM, and SCARA5. Marker genes for alveolar type 1 and type 2 cells included AGER, CAV1, CLDN18, CAV2, ALOX15B, LRRK2, ROS1, SFTPA1, and CSF3R. Marker genes for basal cells included AQP3, KRT5, KRT14, KRT15, ​​TP63, DAPL1, MIR205HG, EYA2, CYP24A1, and KRT17. Marker genes for basophils included CD63, CD203C, CD123, CLC, MS4A3, TCN1, CPA3, HDC, GATA2, MS4A2, IL4, GCSAML, and TPSAB1. Marker genes for B cells and plasma cells included CD79A, CD79B, MS4A1, JCHAIN, TNFRSF17, DERL3, FCRL5, MZB1, VPREB3, IGHA1, IGHM, IGHG1, CD27, SLAMF7, EAF2, and CD38. Marker genes for club cells included SCGB3A2, CYP2B7P, MGP, SFTPC, and SFTPD. Marker genes for pulmonary endothelial cells included CD34, EGFL7, EMCN, FLT1, KDR, TEK, VWF, ACKR1, and CA4.Marker genes for lung fibroblasts included COL3A1, COL5A2, FN1, LUM, LRP1, PDGFRA, TCF21, and SCARA5. Marker genes for goblet cells, serous cells, and mucous cells included MUC5B, MUC5AC, SPDEF, SCGB1A1, SCGB3A1, BPIFB1, LTF, ANPEP, EPCAM, IL10, and LTF. Marker genes for lung ciliated cells included FOXJ1, TUBB1, TP73, CCDC78, and CAPS. Marker genes for lung mesothelial cells included MSLN, UPK3B, WT1, CALB2, VCAM1, MEDAG, GAS1, HAS1, C1S, C2, and CFB. Marker genes for lung monocytes, dendritic cells, and macrophages included CD4, MS4A7, VCAN, FCN1, FCGR3A, ITGAM, CD1C, ITGAX, LILRB2, and CD14. Marker genes for lung neutrophils included S100A8, S100A12, FCGR3B, LILRA5, and G0S2. Marker genes for lung pericytes included KCNK3, CDH6, COX4I2, NDUFA4L2, PDGFRB, CSPG4, TRPC6, and RGS5. Marker genes for lung ionocytes included CFTR, FOXI1, ASCL3, ATP6V1G3, BSND, and HEPACAM2. Marker genes for lung smooth muscle cells included PLN, DES, TNNT2, ACTG2, ATP1A2, MYH11, TAGLN, ACTA2, CALD1, and NOTCH3. Marker genes for lung T cells and natural killer cells included CD4, CD3D, CD3E, CD3G, IL7R, CD8A, CD8B, FOXP3, TIGIT, CD27, NCAM1, KLRF1, GNLY, NKG7, and TNFRSF18.

[0086] As shown in Figures 13A and 13B, the cell type identification method was able to achieve a high similarity of 96% with the manual annotation of the Tabula test for all cell types. Only 14 cells remained unlabeled, and no strong features were observed among the enumerated cell types (Figure 13B). As shown in Figure 13C, the high assignment similarity between the cell type identification method and the manual annotation was maintained at the individual cell type level.

[0087] Example 3: Computational Efficiency Due to the large size of typical scRNA-seq datasets, increasing computational efficiency is advantageous for cell type identification methods. The processing time for the cell type identification method disclosed herein was evaluated using a single core with a 3.1 GHz processor and 128 GB memory. The annotation of 15,000 simulated cells (the largest simulated data set used in this study) using the cell type identification method disclosed herein took less than 2 minutes. Furthermore, the annotation of approximately 27 k ± 20 k (mean ± standard deviation) experimental cells (the average data size used in this study) took approximately 2.3 ± 1.7 minutes.

[0088] Other embodiments While the present methods and compositions have been described in conjunction with the detailed description thereof, it should be understood that the foregoing description is intended to be illustrative, and not limiting, of the scope of the present methods and compositions, as defined by the scope of the appended claims. Other aspects, advantages, and modifications are within the scope of the following claims.

Claims

1. 1. A method for identifying the cell type origin of each cell of a plurality of cells, comprising: receiving a plurality of sequencing read counts; providing a set of cell types to be detected and an associated gene set G for each cell type, such that each gene set G includes a plurality of genes g; Scoring the sequencing read counts to generate an assignment score; assigning each cell of the plurality of cells to a cell type having a highest assignment score to identify the cell type origin of each cell; A method comprising:

2. 10. The method of claim 1, further comprising trimming the sequencing read count to remove ambiguous calls.

3. 3. The method of claim 1 or 2, further comprising smoothing the annotation of the sequencing read counts.

4. The method of any one of claims 1 to 3, wherein the plurality of sequencing read counts are generated by RNA sequencing (RNA-seq).

5. 5. The method of claim 1, wherein the plurality of sequencing read counts are generated by single-cell RNA sequencing (scRNA-seq) of each cell in the plurality of cells.

6. 6. The method of claim 1, wherein the scoring step comprises sorting the plurality of genes g of the associated gene set G that exhibit non-zero expression from high to low expression.

7. 7. The method of claim 1, wherein the scoring step further comprises generating a ranked vector of length N from the sorted genes of the relevant gene set.

8. 8. The method of claim 7, wherein the scoring step further comprises converting the ranked vector of length N into a binary array s such that genes g included in a particular gene set G are replaced by 1 or 0.

9. The binary sequence s has the following formula: [Equation 1] The method of claim 8 , wherein the

10. The method of claim 8 or 9, wherein the scoring step further comprises performing a cumulative sum over the binary array s up to each element k of the binary vector.

11. The scoring step sums over all generated successive partial sums from 1 to k to obtain the following formula: [Equation 2] 11. The method of claim 10, further comprising generating an assignment score S for a given cell according to:

12. The method according to any one of claims 8 to 11, wherein the scoring step further comprises performing the steps described in any one of claims 8 to 11 for each cell and gene set of the plurality of cells, for a total of M gene sets.

13. 13. The method of claim 1, wherein the scoring step further comprises converting the input gene x cell expression matrix into a gene set x cell assignment score matrix, wherein the input gene x cell expression matrix comprises [number of genes x number of cells], and the gene set x cell assignment score matrix comprises [number of gene sets x number of cells].

14. Each cell of the plurality of cells has the following formula: max{S ｍ } １≦ｍ≦Ｍ・ 14. The method of claim 12 or 13, wherein the cell type is assigned to the cell type that represents the highest assignment score, represented by:

15. The method of any one of claims 1 to 14, wherein the trimming step further comprises calculating a Gini index value for each cell of the plurality of cells.

16. The trimming step may be performed by adjusting the gene set x cell assignment score matrix using the following formula: ∈［０，１］ The method of any one of claims 1 to 14, further comprising transforming to an exponential distribution according to

17. 17. The method of claim 15 or 16, wherein the trimming step further comprises excluding cells that are outliers and / or have an index less than 0.

5.

18. 18. The method of any one of claims 1 to 17, wherein the smoothing step further comprises applying a k-nearest neighbor (kNN) algorithm to the sequencing read counts.

19. The smoothing step comprises: Identifying a cell with a minimum k nearest neighbors; If more than 50% of the neighbors reach a consensus annotation, annotating the neighboring cells according to the consensus of the neighbors.

20. The method of claim 18, further comprising:

20. 20. The method of any one of claims 1 to 19, further comprising providing a set of negative markers, wherein cells that do not express one or more of the negative markers are rewarded and cells that express one or more of the negative markers are penalized.

21. The method of any one of claims 1 to 20, wherein the method does not include normalizing the plurality of sequencing read counts.

22. The method of any one of claims 1 to 21, wherein the method does not include converting the plurality of sequencing read counts.

23. The method of any one of claims 1 to 22, wherein the method does not include clustering of the plurality of sequencing read counts.

24. The method of any one of claims 1 to 23, wherein the plurality of cells are human cells.

25. 25. The method of any one of claims 1 to 24, wherein the set of cell types to be detected comprises 1, 2, 3, 4, 5, 6, 7, 8, 9 or 10 cell types to be detected.

26. 26. The method of any one of claims 1 to 25, wherein the plurality of genes g comprises 10, 20, 30, 40, 50, 60, 70, 80, 90 or 100 genes.

27. A computer program product tangibly embodied in a computer-readable medium, comprising instructions that, when executed by one or more processors, receiving a plurality of sequencing read counts; receiving a set of cell types to be detected and an associated gene set G for each cell type, such that each gene set G includes a plurality of genes g; Scoring the sequencing read counts to generate an assignment score; assigning each cell of the plurality of cells to the cell type having the highest assignment score to identify the tissue type origin of each cell; 1. A computer program product configured to:

28. 28. The method of claim 27, further comprising instructions that, when executed by the one or more processors, are configured to trim the sequencing read count to remove ambiguous calls.

29. 29. The method of claim 27 or 28, further comprising instructions that, when executed by the one or more processors, are configured to smooth the annotation of the sequencing read counts.

Images