New biomarkers and their application in the prognosis assessment of liver cancer

Abstract

The present invention belongs to the field of biomedical technology, and specifically relates to a novel marker and an application thereof in the prognosis evaluation of liver cancer. The novel marker combination includes HLA.DRB5, CXCL9, ACKR1, CAMK2B, SOWAHD, ZNF30, LINC01428, SERHL2, ADAMTS20 and ITPRIPL1. The present invention screens out a gene set of ten key genes that distinguish between TLS tending to mature and TLS deviating from mature by distinguishing between the two groups, and finally establishes an effective classification model. It has been verified that the model can be effectively applied to predict the clinical prognosis of patients with hepatocellular carcinoma.

Description

Technical Field

[0001] The present invention belongs to the field of biomedical technology, and specifically relates to a novel marker and its application in liver cancer prognosis assessment. Background Art

[0002] Primary hepatocellular carcinoma (HCC) is the digestive system tumor with the highest mortality rate. China ranks first in the world in the number of newly diagnosed liver cancers. The high recurrence rate after liver cancer surgery is the key factor affecting the long-term survival of patients. Data show that the recurrence rate of liver cancer within 5 years after surgery is as high as 70% under local surgical treatment, interventional ablation and other methods, and more than 50% of patients die within 2 years of recurrence. Surgical treatment is a first-line option. According to traditional clinical characteristics, the prognosis of patients of the same type after surgical treatment is also very different. In recent years, studies have shown that this is mainly caused by the molecular heterogeneity of cancer patients. Early screening and early diagnosis and treatment are the first line of defense for liver cancer patients to avoid missing the opportunity for surgery and late chemotherapy resistance and tumor recurrence. Early diagnosis can also identify and distinguish low-risk and high-risk patient populations and predict the prognosis of patients. Therefore, accurate guidance of treatment decisions is the key to improving the efficacy of liver cancer. It is urgent to find powerful and reliable liver cancer prognostic markers.

[0003] Tertiary lymphoid structures (TLS) are ectopic lymphoid aggregates that form in non-lymphoid tissues, including tumors, in response to immune stimulation. TLS of varying maturity states are present in a variety of cancers at all stages of disease. Mature TLS possess germinal centers that are critical for B-cell development and function. In contrast, immature TLS lack well-formed germinal centers and are unable to adequately support the full acquisition of B-cell function. In a variety of cancers, including hepatocellular carcinoma, melanoma, non-small cell liver cancer, and renal cell carcinoma, the presence of mature TLS and B-cell infiltration are positively correlated with improved patient prognosis and enhanced immune response. Therefore, TLS are increasingly being used to predict response to immunotherapy, but in reality these predictions are only valid in a minority of patients. Increasing evidence highlights the critical role of mature TLS as endogenous regulators of antitumor immunity and in enhancing antigen presentation, cytokine-mediated signaling, and the release of tumor-specific antibodies. The frequency of mature TLS varies considerably between cancers and patients, reflecting their dynamic development and complex response to local microenvironmental signals. This heterogeneity of TLS may affect the overall anti-tumor immune response and clinical prognosis. Despite these insights, accurately determining the developmental status of TLS remains a huge challenge due to the lack of a standardized classification system applicable to different TLS populations. In addition, the factors that promote TLS maturation within the tumor remain unknown, and there is currently no strategy to improve the intratumor immune response by promoting TLS maturation. Identifying the developmental stage of TLS is a research hotspot in the field of tumor immunology, and finding key markers for identifying TLS subtypes is of great significance for predicting the population that may benefit. Summary of the invention

[0004] The main purpose of the present invention is to provide a new type of marker and its application in the prognosis assessment of liver cancer. By selecting a reasonable method to screen out relevant gene molecular markers and clarify the potential molecular mechanism of liver cancer, it can provide direction for the prognosis prediction and diagnosis of liver cancer and further improve the clinical diagnostic efficiency and prognosis judgment ability of liver cancer patients.

[0005] To achieve the above objectives, the present invention provides the following technical solutions:

[0006] A marker combination, the marker combination comprising HLA.DRB5, CXCL9, ACKR1, CAMK2B, SOWAHD, ZNF30, LINC01428, SERHL2, ADAMTS20 and ITPRIPL1.

[0007] Furthermore, the present invention provides an application of the marker combination, which is any one of the following a1) to a7);

[0008] a1) Preparation of products for prognostic risk assessment of liver cancer patients;

[0009] a2) preparing a product for stratifying the prognosis of liver cancer patients;

[0010] a3) Preparation of products for enrichment of prognosis of liver cancer patients

[0011] a4) preparing a product for predicting the survival rate of liver cancer patients;

[0012] a5) Preparation of products for evaluating the prognosis and efficacy of medication for patients with liver cancer;

[0013] a6) Construct a model / system / device / computer program product for liver cancer prognosis risk assessment;

[0014] a7) Construct a model / system / device / computer program product for predicting the prognosis and survival rate of liver cancer.

[0015] As an embodiment, the present invention provides a product for liver cancer prognosis assessment or diagnosis, wherein the product contains a reagent for detecting the expression information of the marker combination.

[0016] Furthermore, the product is an in vitro diagnostic product.

[0017] Furthermore, the product includes a test kit, a test paper or a gene chip.

[0018] Furthermore, the reagents include reagents for detecting the expression information of the marker combination in the subject's sample at the gene level and / or protein level.

[0019] Further, the reagents include reagents for detecting the expression information of the marker combination in the subject sample by immunohistochemistry, in situ molecular hybridization histochemistry, Western blotting, Northern blotting, PCR, RT-PCR or biochip detection technology;

[0020] Preferably, the immunohistochemistry method includes: immunofluorescence analysis, reverse enzyme-linked immunosorbent assay or immunocolloidal gold method.

[0021] Furthermore, the sample is from the subject's blood, tissue, cell sample, urine or feces.

[0022] Furthermore, the subject is a mammal, preferably, the subject includes mice, rabbits, pigs, non-human primates or humans.

[0023] As a specific embodiment, the present invention provides a liver cancer prognosis risk assessment model, which is a random forest model, and the model collects expression information of each marker in the marker combination.

[0024] Furthermore, the method for constructing the model comprises the following steps:

[0025] S1: Subject samples were collected and spatial transcriptome sequencing was performed;

[0026] S2: Identification of TLS based on spatial transcriptome data;

[0027] S3: Differentiation between mature and immature TLS based on spatial transcriptome data, transcriptome data of all naive B and germinal center B cells in the identified mature TLS;

[0028] S4: Construct a TLS maturation trajectory with naive B cells in mature TLS as the starting point and germinal center B cells as the end point;

[0029] S5: Methods for identifying TLS that is moving toward maturity and TLS that is deviating from maturity:

[0030] All naive B cells and germinal center B cells in the immature TLS to be identified are included and mapped to the TLS maturation trajectory. A regression equation is constructed to predict the development trend of TLS based on the slope of the linear regression. The slope fits the transcriptional similarity of naive B cells and germinal center B cells along the maturation trajectory in each immature TLS. According to the trend of the slope of the linear fitting equation of each immature TLS to be identified, the TLS in the positive slope group is classified as TLS tending towards maturity, and the TLS in the negative slope group is classified as TLS deviating from maturity.

[0031] S6: Identify TLS tending to mature and TLS deviating from mature, collect the expression information of each marker in the marker combination of the two TLS, use the random forest method for training and verification, and the constructed random forest model is the liver cancer prognosis risk assessment model.

[0032] Preferably, in step S3, TLS with a CR2-positive B cell ratio of not less than 5% of all cells in the TLS are defined as mature TLS, and for the remaining TLS, TLS with less than 5 naive B cells and germinal center B cells are filtered out, and TLS with more than 5 naive B cells and GC-B cells are classified as immature TLS;

[0033] Preferably, in step S4, the method for constructing the TLS maturity trajectory is:

[0034] S 4-1: After the transcriptome data were standardized and normalized, the genes related to the mature TLS co-expression module were used to perform PCA analysis on TLS by the Hotspot toolkit. The first 50 principal components were obtained from the PCA analysis for subsequent batch correction by the Harmony toolkit;

[0035] S4-2: Harmony's coordinates were then input into the SPRING toolkit to construct a k-nearest neighbor (kNN) network, and the Monocle3 toolkit was subsequently used to perform a two-dimensional visualization of the trajectory of naive B to germinal center B cells. Based on the coordinates of the kNN network, the learn_graph() function in the Monocle3 package was used to represent the cells on a pseudo-time trajectory;

[0036] S4-3: Use the order_cells() function in the Monocle3 package to assign a pseudo-time value to each cell along the trajectory from naive B to germinal center B cells. Finally, apply the slingshot() function in the Slingshot toolkit to fit a smooth trajectory, which is the TLS maturation trajectory.

[0037] Preferably, in step S5, the regression equation is constructed by:

[0038] S5-1: The TLS maturation trajectory is evenly divided into several bins. In each bin, the average expression of all cells is taken to calculate the gene expression profile and create a pseudo batch of cells;

[0039] S5-2: Then, the gene expression profiles associated with mature TLS were extracted to represent the reference cell state at each pseudo-time point in the pseudo-batch data and referred to as “reference cells”;

[0040] S5-3: Calculate the Spearman rank correlation between each B cell to be identified from the immature TLS and each reference cell, and align them to the most appropriate reference cell based on the maximum value of the correlation between the B cell to be identified and the reference cell;

[0041] S5-4: After comparing all the B cells to be identified with the reference cells, the least squares method was used to fit the regression line based on the pseudo time points and correlation scores: the slope of the regression line represents the overall trend along the TLS maturation trajectory. Based on the linear regression results, the Mfuzz toolkit was used to divide the immature TLS into two different groups. According to the positive or negative slope of the linear regression results, the TLS with a positive slope was classified as TLS tending towards maturity, while the TLS with a negative slope was classified as TLS deviating from maturity. The B cells to be identified were naive B cells or germinal center B cells.

[0042] Furthermore, the TLS tending to be mature and the TLS deviating from maturity are randomly split into training set and validation set in the ratio of 8:2.

[0043] As another more specific embodiment, the present invention provides a system for liver cancer prognosis risk assessment, comprising the following modules:

[0044] a) Data collection module: collect samples from subjects, identify samples containing immature TLS, collect expression information of each marker in the marker combination described in the immature TLS samples, and output the expression information data to the model calculation module;

[0045] b) Model calculation module: used to provide the data obtained by the data collection module as input data to any of the trained evaluation models;

[0046] c) Output prediction module: predict the prognosis of the subject based on the predicted probability of the subject obtained by the evaluation model as the risk score of the subject; compare the risk score with the threshold, and the prognosis of the subject below the threshold is better than the prognosis of the subject above the threshold.

[0047] Furthermore, in the output prediction module, the threshold is 0.63.

[0048] As an embodiment, the present invention further provides a computer device / apparatus, the computer device comprising a memory and a processor, the memory storing a program, and the processor implementing the following method when executing the program:

[0049] S1: Obtaining expression information data of each marker in the marker combination in the sample containing immature TLS;

[0050] S2: providing the data as input data to any of the trained evaluation models; and outputting the prognosis result of the subject.

[0051] The computer equipment of the present invention includes (but is not limited to) any personal computer, server or other terminal that can interact with the user through a keyboard, touchpad or voice control device. The computing device herein may also include a mobile terminal, which includes (but is not limited to) any electronic device that can interact with the user through a keyboard, touchpad or voice control device, such as a tablet computer, a smart phone, a personal digital assistant (PDA), a smart wearable device or other terminal. The network in which the computing device is located includes (but is not limited to) the Internet, a wide area network, a metropolitan area network, a local area network, a virtual private network (VPN), etc.

[0052] The memory includes nonvolatile and / or volatile memory. The nonvolatile memory may include a read-only memory (ROM), a programmable ROM (PROM), an electrically programmable ROM (EPROM), an electrically erasable programmable ROM (EEPROM) or a flash memory. The operating system code is stored thereon. For example, the memory also stores codes or instructions, and by running these codes or instructions, the liver cancer prognosis risk assessment model provided by the embodiments disclosed herein can be implemented. The volatile memory may include a random access memory (RAM) or an external cache memory.

[0053] The processor may include one or more microprocessors, digital processors. The processor may call the memory

[0054] The program code stored in the memory is used to execute related functions. The processor is also called a central processing unit (CPU), which can be a large-scale integrated circuit, a computing core (Core) and a control core (Control Unit).

[0055] As another more specific embodiment, the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium is used to store a computer program / instruction; when the computer program / instruction is executed, the following method is implemented:

[0056] S1: Obtain the expression information data of each marker in the marker combination in the sample containing immature TLS;

[0057] S2: providing the data as input data to any of the trained evaluation models; and outputting the prognosis result of the subject.

[0058] As another more specific embodiment, the present invention provides a computer program product, including a computer program / instruction, which implements the following method when executed:

[0059] S1: Obtaining expression information data of each marker in the marker combination in the sample containing immature TLS;

[0060] S2: providing the data as input data to any of the trained evaluation models; and outputting the prognosis result of the subject.

[0061] It will be appreciated by those skilled in the art that all or some of the steps and systems in the methods disclosed above may be implemented as software, firmware, hardware, and appropriate combinations thereof. Some or all physical components may be implemented as software executed by a processor, such as a central processing unit, a digital signal processor, or a microprocessor, or as hardware, or as an integrated circuit, such as an application specific integrated circuit. Such software may be distributed on a computer-readable medium, which may include a computer storage medium (or a non-transitory medium) and a communication medium (or a transient medium).

[0062] The technical effects achieved by the present invention are:

[0063] The present invention proposes a new classification strategy, which divides typical immature TLS into three groups: mature type, maturation-oriented type and deviated from maturity type according to their developmental trajectory relative to the TLS maturation process, characterizes TLS types with different developmental states, and identifies key markers of TLS subtypes. In order to fully realize these values, a marker combination including ten genes such as CXCL9 is designed to accurately predict the clinical prognosis of HCC patients with classic immature TLS. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 This is a diagram of the maturity classification process of TLS in liver cancer;

[0065] A is the in situ image of H&E staining, cell annotation and CR2 gene expression of mature TLS and immature TLS; B is the differentiation trajectory of mature TLS constructed based on the gene expression profile of naive B cells and germinal center B cells in mature TLS; C shows the classification of individual TLS into different maturity types based on the similarity mapping of the differentiation trajectory of naive B cells and germinal center B cells of TLS with mature TLS.

[0066] Figure 2 This is a diagram of the screening process of TLS-related liver cancer marker combination;

[0067] A is the ROC curve of the liver cancer training set, which shows the accuracy of different models in distinguishing between TLS tending to mature and TLS deviating from maturity in the training set; B is the ROC curve of the liver cancer validation set, which shows the accuracy of different models in distinguishing between TLS tending to mature and TLS deviating from maturity in the validation set; C is the ROC curve of the liver cancer validation set, which shows the accuracy of the classifier using the ten characteristic genes screened by the random forest model to distinguish between TLS tending to mature and TLS deviating from maturity.

[0068] Figure 3 This is a graph showing the prognostic effects of ten characteristic genes on liver cancer patients;

[0069] A is the prognosis of liver cancer patients from the First Affiliated Hospital of Sun Yat-sen University; B is the prognosis of liver cancer patients from the TCGA database. DETAILED DESCRIPTION

[0070] The following will clearly and completely describe the concept of the present application and the technical effects produced in combination with the embodiments, so as to fully understand the purpose, features and effects of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of them. Based on the embodiments of the present application, other embodiments obtained by those skilled in the art without creative work are all within the scope of protection of the present application.

[0071] For commodities purchased in the test method, if the specific conditions are not specified, the test shall be carried out under conventional conditions or the conditions recommended by the manufacturer. If the manufacturer of the reagents or instruments used is not specified, conventional products can be purchased from the market.

[0072] Unless otherwise defined herein, scientific and technical terms used in conjunction with the present disclosure shall have the meanings commonly understood by those of ordinary skill in the art. Exemplary methods and materials are described below, but methods and materials similar or equivalent to those described herein can also be used in the practice and testing of the present disclosure.

[0073] As used herein, the following words and terms shall have the meanings indicated:

[0074] The term "subject" refers to mammals, including mice, rabbits, non-human primates (monkeys, etc.) and / or humans, etc. "Murine" refers to any mammal from the family Muridae, such as mice, rats, etc.

[0075] The term "threshold value" refers to a value determined for a specific variable value, wherein the value gives a certain meaning when the change value is greater than or less than the value.

[0076] The term "marker" refers to an indicator that can be objectively measured and evaluated, reflects physiological or pathological processes, and produces biological effects in response to exposure or therapeutic interventions. In the field of oncology, biomarkers are usually biological substances produced by tumor cells or non-tumor cells that reflect the presence and changes of tumor cells or non-tumor cells in the body.

[0077] The term "subject stratification" refers to the fact that prognostic markers can distinguish between patient populations with the same diagnosis but different natural disease courses who have not received therapeutic intervention. Prognostic markers are used in controlled studies to stratify subjects according to differences in prognosis, thereby reducing the interference of subject heterogeneity and confounding factors on the test results, reducing inter-group bias, and improving the reliability of the results.

[0078] The term "enriched population" refers to the fact that prognostic markers can help screen out patient populations that are more likely to have the clinical events of interest. Therefore, in clinical trials with clinical events as the endpoint, the use of prognostic markers to screen enriched high-risk patient populations, such as selecting high-risk patient populations with high recurrence rates and poor prognosis for research, can reduce the sample size of clinical trials and more efficiently obtain the number of clinical events sufficient to support evaluation. Choosing to conduct research in high-risk enriched populations does not mean that low-risk patient populations with the same indication cannot benefit from treatment; after successful clinical trials in high-risk populations, further consideration can be given to expanding research to appropriate low-risk patients.

[0079] The term "prognosis" refers to the prediction of the possibility, development process and outcome of various diseases, and can determine some consequences of the disease, such as recovery, recovery, or even no recovery, improvement, death, etc.

[0080] The term "sample" refers to a sample obtained directly from a source of interest by any appropriate means, such as selected from a method selected from biopsy (eg, fine needle aspiration or tissue biopsy), surgery, collection of body fluids (eg, blood, lymph, etc.), and the like.

[0081] The term "TLS-50 gene set" is named TLS-50signature in English. Reference: Rui W, Wenbo G, Xinyao Q, et al. Comprehensive analysis of spatial architecture in primary liver cancer. [J]. Science advances, 2021, 7(51): eabg3750-eabg3750. This study identified the TLS region in the spatial transcriptome data by integrating 10X Visium spatial transcriptome data with H&E staining, and used the top 50 genes (TLS-50) with high expression in this region as the signature genes of TLS.

[0082] The term "12-chemokine signature gene set" is named in English as 12chemokine gene subset. Reference: Coppola D, Nebozhyn M, Khalil F, et al. Unique Ectopic Lymph Node-LikeStructures Present in Human Primary Colorectal Carcinoma Are Identified by Immune Gene Array Profiling [J]. The American Journal of Pathology, 2011, 179(1): 37-45. This study used the gene expression data of 326 colorectal cancer patients to screen out a gene set containing 12 chemokine genes and found that TLS was significantly correlated with the 12-chemokine gene expression signature, suggesting that if the patient's tumor tissue highly expresses these 12 chemokines, TLS is more likely to be present in the tumor tissue.

[0083] The term "naive B cells ( "B Cell") is a type of B lymphocyte that has not been immune stimulated and has not yet undergone antigen activation. They are usually found in lymphoid organs (such as lymph nodes, spleen) and peripheral blood. Naive B cells express B cell receptors (BCR), but these receptors have not yet been exposed to the corresponding antigens. When naive B cells bind to their specific antigens, they will be activated and begin to differentiate, transforming into germinal center B cells or memory B cells, thereby participating in the response of humoral immunity. Usually, naive B cells have lower expression levels because they have not yet experienced antigen selection and stimulation, and therefore lack memory for specific antigens. Naive B cells usually highly express genes such as IGHD, TCL1A, CD22, and CD72.

[0084] The term "germinal center B cells" refers to a specific type of B lymphocytes in lymph nodes, spleen and other lymphoid organs. They usually appear in the germinal center of the lymph node. The germinal center is a specific area in the lymph node reaction that is used to promote the proliferation, differentiation and affinity maturation of B cells in the immune response. Germinal center B cells have highly active metabolism, DNA recombination and highly expressed surface molecules, such as highly expressed AICDA, MKI67, BCL6, RGS13 and other genes. They play a key role in the affinity maturation of antibodies and the generation of memory B cells.

[0085] The marker information protected in the present invention is shown in Table 1 below:

[0086] Table 1

[0087]

[0088]

[0089] In this study, we used Stereo-seq, a large-field spatial transcriptomics method with single-cell resolution, to comprehensively map TLS in situ and analyze its microenvironment, and comprehensively demonstrated the spatial characteristics of TLS in hepatocellular carcinoma (HCCs). We also proposed a new classification strategy, which divides typical immature TLS into mature-type and deviated-mature TLS according to their developmental trajectory relative to the TLS maturation process, revealing the unique characteristics of mature-type TLS in anti-tumor immunity, and characterizing TLS types with different developmental states, and identifying key markers for TLS subtypes.

[0090] Example 1: Construction of a liver cancer combined molecular marker model

[0091] 1. Spatial transcriptome sequencing and TLS identification based on liver cancer specimens

[0092] 1.1. Collection of liver cancer pathological specimens

[0093] Ten patients with hepatocellular carcinoma who underwent surgical treatment from the First Affiliated Hospital of Sun Yat-sen University (SYSU-FAH) were included. Inclusion criteria: met the diagnostic criteria for hepatocellular carcinoma in the "Guidelines for the Diagnosis and Treatment of Primary Liver Cancer (2019 Edition)"; did not receive immunosuppressive therapy or anti-tumor targeted therapy before surgery. Exclusion criteria: combined with cholangiocarcinoma or other malignant tumors. Surgical resection specimens of the above patients were collected, including tumor tissue, junctional tissue, and paracancerous liver tissue. The junctional tissue was the tissue with a width of 0.5 cm from the tumor invasion edge to the inside and outside. The paracancerous liver tissue was the liver tissue at least 2 cm away from the tumor invasion edge. When sampling, ischemic necrotic tissue was avoided as much as possible. The fresh tissue samples were placed on ice within 30 minutes of ex vivo ex vivo, trimmed into tissue blocks with a length × width <10 mm × 10 mm, 3 mm ≤ height ≤ 8 mm, and then placed horizontally in a tissue embedding box filled with OCT embedding agent, and then the latter was placed on the surface of liquid nitrogen for quick freezing until the embedding agent became hard and white, and finally stored in a -80℃ refrigerator. In addition, the target samples need to be sliced ​​and RNA quality assessed, requiring intact tissue, clear embedding direction, and RNA integrity RIN ≥ 7. The sample collection process is reviewed by the hospital ethics committee and complies with international bioethical standards.

[0094] 1.2 Spatial transcriptome sequencing

[0095] Each tissue sample is sequenced using BGI Stereo-seq spatial transcriptome technology (each detection point has a diameter of 220nm, and the center distance between two detection points is 500nm). The specific process is as follows: slice the OCT-embedded sample, and use fluorescence imaging technology to microscopically observe the target tissue slices. The fluorescence imaging results can confirm the basic information such as the shape contour, tissue structure, and completeness of the target sample tissue. The staining effect requires that the cell resolution rate is ≥90%. The target slice is then attached to the surface of the sequencing chip, and the tissue attached to the chip is sliced, fixed, and infiltrated. The chip will capture the mRNA molecules of the target sample tissue cells and synthesize cDNA to construct a sequencing library, and finally the sequencing library is sequenced.

[0096] 1.3 TLS identification based on spatial transcriptome data

[0097] After quality control, alignment, and quantification of the original FASTQ sequence (for the specific implementation of this process, refer to https: / / github.com / BGIResearch / SAW), the expression spectrum matrix that retains the spatial position information of each sample is obtained, and downstream analysis is performed using a resolution of Bin50 (50×50 DNA nanospheres / Bin, 25 μm in length).

[0098] Then, TLS was identified in situ on the spatial transcriptome data. The specific process is as follows:

[0099] 1) Pathological identification: Pathologists marked TLS candidate regions on H&E stained sections. This manual annotation provided additional evidence for the subsequent confirmation of the presence of TLS in spatial transcriptome data;

[0100] 2) Signature gene sets associated with TLS: Two different TLS signature gene sets, namely the TLS-50 gene set and the 12-chemokine signature gene set, were calculated using the AddModuleScore() function of the Seurat toolkit (v4.0.5). These signature gene sets showed spatial expression patterns associated with TLS.

[0101] 3) Spatial co-expression modules associated with TLS: The Hotspot toolkit (v1.1) was used to identify regions with characteristic spatial co-expression patterns of TLS.

[0102] 4) Identification of immune cell clusters: Use the NormalizeData(), ScaleData(), RunPCA(), FindNeighbors(), FindClusters(), and RunUMAP() functions of the Seurat toolkit (v4.0.5) and the SCTransform() function in the sctransform toolkit (v0.3.2) to reduce the dimension and cluster the spatial transcriptome data to obtain cluster subgroups (cluster). Then use the FindAllMarkers() function to calculate the characteristic genes of each cluster. Based on the characteristic genes, identify clusters with immune cell characteristics and define the latter as "immune cell clusters". Use the OPTICS algorithm to identify immune cell clusters with high immune cell density among the previously found immune cell clusters.

[0103] 5) TLS region identification: The presence of immune cell clusters was supported by most of the signatures in these regions (gene signatures, co-expression modules, and H&E staining). The OPTICS parameters were fine-tuned to maximize TLS identification while taking these criteria into account, and the closing operation in OpenCV (v4.6.0.66) was used to fill in the holes within the TLS regions.

[0104] 6) Optimize the TLS area: Based on the differences in cell density observed in the staining images, morphological operations were applied to adjust the TLS boundaries and further optimize the TLS area dominated by non-immune cells.

[0105] 7) Filtering and analysis: TLS with an area less than 20 Bin50 were filtered out, and the remaining TLS were used for downstream maturity typing analysis.

[0106] 2. Maturity classification of TLS in liver cancer

[0107] 2.1. Distinguishing between mature and immature TLS in in situ spatial transcriptome data

[0108] TLS with a CR2-positive B cell ratio of no less than 5% of all cells in the TLS were defined as mature TLS. For the remaining TLS, naive B cells and TLS with less than 5 germinal center B cells were filtered out, and the remaining TLS after filtering were classified as immature TLS ( Figure 1 (A).

[0109] 2.2. Constructing a Pseudo-Timing Trajectory of Mature TLS

[0110] Taking naive B cells in mature TLS as the starting point and germinal center B cells as the end point, a TLS maturation trajectory can be constructed to simulate the process of TLS maturation at the transcriptome level. First, the transcriptome data of all naive B and germinal center B cells in mature TLS identified from 2.1 were included. After standardizing and normalizing the data, we performed principal component analysis (PCA) on 1,118 genes associated with TLS co-expression modules (identified by Hotspot toolkit (v1.1)). The first 50 principal components (PCs) obtained from the PCA analysis were used for subsequent batch correction by Harmony toolkit (v0.1.1).

[0111] Harmony's coordinates were then fed into the SPRING toolkit to build a k-nearest neighbor (kNN) network ( Figure 1 B), and then the trajectory of naive B to germinal center B cells was visualized in two dimensions using the Monocle3 toolkit (v1.0.0). Based on the coordinates of the kNN network, we used the learn_graph() function in the Monocle3 package to represent the cells on the pseudo-time trajectory. Then, we used the order_cells() function in the Monocle3 package to assign a pseudo-time value to each cell along the trajectory from naive B to germinal center B cells. Finally, we applied the slingshot() function in the Slingshot toolkit (v2.6.0) to fit a smooth trajectory, which is therefore called the reference trajectory during TLS maturation. The above analysis was completed in the R programming language (version 4.1.0).

[0112] 2.3. Further incorporate naive B and germinal center B cells in each TLS and map them to the TLS maturation trajectory. The next goal is to predict the development trend of TLS based on the slope of the linear regression, which simulates the transcriptional similarity of naive B and germinal center B cells along the maturation trajectory in each immature TLS. First, the reference TLS maturation trajectory is evenly divided into 100 segments. Within each segment, the average expression of all cells is taken to calculate the gene expression profile to create a pseudo-bulk of cells. Then, the 1,118 TLS-related gene expression profiles in 2.2 are extracted to represent the reference cell state at each pseudo-time point in the pseudo-bulk data and are referred to as "reference cells". Next, the Spearman rank correlation between each B cell to be identified from the TLS (naive B or germinal center B cell) and each reference cell is calculated. By aligning the B cell to be identified with the reference cell with the highest correlation, its most likely cell state is determined. This alignment process is used for all B cells to align them with the reference cells along the maturation trajectory. After aligning all the B cells to be identified with the reference cells, the least squares method was applied to fit the regression line based on the pseudo time points and the correlation scores. The slope of the regression line represents the overall trend along the TLS maturation trajectory. Based on the linear fitting equation of each TLS (y = βx + α, each TLS has its own intercept α and slope β), we used the Mfuzz toolkit (v2.58.0) to cluster all immature TLS into two groups. According to the trend of the slope of the linear fitting equation of all TLS in each group, the TLS in the positive slope group was named as TLS tending to mature, and the TLS in the negative slope group was named as TLS deviating from mature ( Figure 1 The above analysis was completed in R programming language (version 4.1.0).

[0113] 3. Discovery of a combination of ten liver cancer markers

[0114] In order to find markers that can distinguish TLS tending to mature and those deviating from maturity, all TLS tending to mature and those deviating from maturity (a total of 258) were randomly split into training set and validation set at a ratio of 8:2. A model for predicting TLS deviating from maturity was established in the training set, and the model effect was verified in the validation set. First, the gene expression profiles of TLS tending to mature and those deviating from maturity in the training set (a total of 22,240 genes) were screened using the following rules:

[0115] (1) The p value in the difference test is less than 0.05;

[0116] (2) the fold change between the two groups was greater than 2;

[0117] (3) The AUC of the gene is greater than 0.6.

[0118] Finally, 90 genes were left for subsequent modeling. The following machine learning methods were used to build models, including: Lasso-logistic, Decision Tree, Random Forest, Support Vector Machine, NaiveBayes and XGBoost ( Figure 2 The best model in the validation set, i.e. the model built using the Random Forest method for the top 10 genes: HLA.DRB5, CXCL9, ACKR1, CAMK2B, SOWAHD, ZNF30, LINC01428, SERHL2, ADAMTS20, and ITPRIPL1, was selected as the final model ( Figure 2 In B), the area under the ROC curve (AUC) of this effective classification model in the validation set is 0.8 ( Figure 2 (C) shows that the combination of these ten markers can accurately distinguish TLS that tend to mature from those that deviate from maturity in the validation set.

[0119] Example 2 Judgment of liver cancer prognosis

[0120] 4.1 Differentiating between mature and immature TLS

[0121] 4.1.1 Cases and samples

[0122] A total of 422 patients with hepatocellular carcinoma who underwent surgical treatment at the First Affiliated Hospital of Sun Yat-sen University (SYSU-FAH) were included. Inclusion criteria: met the diagnostic criteria for hepatocellular carcinoma in the "Guidelines for the Diagnosis and Treatment of Primary Liver Cancer (2019 Edition)"; did not receive immunosuppressive therapy or anti-tumor targeted therapy before surgery; had available postoperative pathological sections including tumors. Exclusion criteria: combined with cholangiocarcinoma or other malignant tumors; incomplete follow-up data. Hematoxylin-eosin (H&E)-stained pathological sections and tumor tissue transcriptome sequencing data of these patients were collected. Clinical and follow-up information was obtained from the case database of the First Affiliated Hospital of Sun Yat-sen University.

[0123] 4.1.2 TLS Identification and Maturity Assessment

[0124] When HE staining results were observed under an ordinary optical microscope, patients with at least one lymphocyte cluster structure in the tumor area were defined as TLS-positive patients, otherwise they were defined as negative. Subsequently, TLS can be divided into two levels according to its degree of aggregation and the presence or absence of germinal centers: immature TLS: lymphocyte clusters with vague and slightly sparse immune cell aggregation; mature TLS: obvious lymphocyte clusters (TLS cells are dense and TLS shape is round) with or without light-stained germinal center formation. Among TLS-positive patients, if all TLS are immature TLS, the patient is divided into the patient cohort of immature TLS; if there is at least one mature TLS, the patient is divided into the patient cohort of mature TLS.

[0125] 4.2 Calculation of comprehensive scores of combined molecular markers

[0126] The random forest model established with the above ten genes was used to predict the tumor tissue transcriptome sequencing dataset of the SYSU-FAH immature TLS patient cohort (155 cases). The expression information of the ten genes in the dataset was collected and input into the model to obtain the prediction probability, and the prediction probability of the model was used as the patient's risk score.

[0127] 4.3 Obtaining the Threshold

[0128] In the tumor tissue transcriptome sequencing dataset of the SYSU-FAH immature TLS patient cohort (155 cases), the "maxstat": maximally selected rank statistics method was used to select the optimal threshold of the risk score as 0.63 for the prognostic outcome RFS (recurrence-free survival). These patients were grouped according to this threshold. Patients with a value greater than this threshold were classified as the low group of the mature TLS gene set, and patients with a value less than this threshold were classified as the high group of the mature TLS gene set.

[0129] 4.4 Combination of molecular markers for predicting the prognosis of patients with liver cancer

[0130] According to the above method, the FAH-SYSU patient cohort was divided into mature TLS group, mature TLS gene set high group, mature TLS gene set low group and TLS negative group. Kaplan-Meier survival analysis showed that the prognosis of patients in the mature TLS gene set high group was significantly better than that in the mature TLS gene set low group (p<0.0001, Figure 3 (A).

[0131] Example 3 Combination of molecular markers for prognosis prediction of liver cancer patients in TCGA public database

[0132] 5.1 Case collection and TLS identification

[0133] A total of 370 cases of hepatocellular carcinoma with no concomitant tumors, complete clinical data, and available postoperative HE pathological sections including tumors were collected from the Cancer Genome Atlas (TCGA) open source database (https: / / portal.gdc.cancer.gov). The 370 patients were then divided into mature TLS group, immature TLS group, and TLS negative group according to the steps in 4.1.

[0134] 5.2 Prognosis prediction

[0135] The random forest model established with the above ten genes was used to predict the tumor tissue transcriptome sequencing data set of the TCGA immature TLS patient cohort (125 cases). The model will use the expression information of the above 10 genes in the transcriptome data for prediction, and the prediction probability of the model will be used as the patient's risk score. The patients were then grouped according to the cutoff value obtained in 4.3. Patients with a p-value greater than 0.63 were classified as the low group of the mature TLS gene set, and patients with a p-value less than 0.63 were classified as the high group of the mature TLS gene set. Kaplan-Meier survival analysis showed that the prognosis of patients in the high group of the mature TLS gene set was significantly better than that of patients in the low group of the mature TLS gene set (p=0.0021, Figure 3 (middle B).

[0136] In summary, the present invention distinguishes between mature-type and mature-deviant TLS and explores their role in predicting the clinical prognosis of patients with hepatocellular carcinoma. After screening, a gene set of ten key genes that distinguish these two groups was listed, and finally an effective classification model (AUC = 0.80, Figure 2 C). Using this model, we divided patients with only classical immature TLS into two groups (high and low groups with a trend toward mature TLS gene set). Notably, patients in the high group, similar to the mature TLS group, showed significantly longer recurrence-free survival (RFS) in both the internal validation cohort (p<0.0001) and the external TCGA cohort (p=0.0021) ( Figure 3 Our results suggest that patients with immature TLS identified by a marker panel constructed based on the distinguishing features we found between mature-type and mature-deviant TLS can be further stratified or population enriched to predict their clinical prognosis.

Claims

1. A liver cancer prognostic risk assessment model. It is characterized in that The model is a random forest model, which collects expression information of each marker in the marker combination, and the marker combination is composed of HLA.DRB5, CXCL9, ACKR1, CAMK2B, SOWAHD, ZNF30, LINC01428, SERHL2, ADAMTS20 and ITPRIPL1 genes.

2. The model according to claim 1, It is characterized in that The method for constructing the model comprises the following steps: S1: Subject samples were collected and spatial transcriptome sequencing was performed; S2: Identification of TLS based on spatial transcriptome data; S3: Differentiation between mature and immature TLS based on spatial transcriptome data, transcriptome data of all naive B cells and germinal center B cells in the identified mature TLS; S4: Construct a TLS maturation trajectory with naive B cells in mature TLS as the starting point and germinal center B cells as the end point; S5: Methods for identifying TLS that is moving toward maturity and TLS that is deviating from maturity: All naive B cells and germinal center B cells in the immature TLS to be identified are included and mapped to the TLS maturation trajectory. A regression equation is constructed to predict the development trend of TLS based on the slope of the linear regression. The slope fits the transcriptional similarity of naive B cells and germinal center B cells along the maturation trajectory in each immature TLS. According to the trend of the slope of the linear fitting equation of each immature TLS to be identified, the TLS in the positive slope group is classified as TLS tending towards maturity, and the TLS in the negative slope group is classified as TLS deviating from maturity. S6: Identify TLS tending to mature and TLS deviating from mature, collect the expression information of each marker in the marker combination of the two TLS, use the random forest method for training and verification, and the constructed random forest model is the liver cancer prognosis risk assessment model.

3. The model according to claim 2, It is characterized in that In step S3, TLS with a CR2-positive B cell ratio of not less than 5% of all cells in the TLS were defined as mature TLS, and for the remaining TLS, TLS with less than 5 naive B cells and germinal center B cells were filtered out, and those with more than 5 naive B cells and GC-B cells were classified as immature TLS.

4. The model according to claim 2, It is characterized in that In step S4, the TLS maturity trajectory is constructed as follows: S4-1: After the transcriptome data were standardized and normalized, the genes related to the mature TLS co-expression module were used to perform PCA analysis on TLS by the Hotspot toolkit. The first 50 principal components were obtained from the PCA analysis for subsequent batch correction by the Harmony toolkit. S4-2: Harmony's coordinates were then input into the SPRING toolkit to construct a k-nearest neighbor (kNN) network, and the Monocle3 toolkit was subsequently used to perform a two-dimensional visualization of the trajectory of naive B cells to germinal center B cells. Based on the coordinates of the kNN network, the learn_graph() function in the Monocle3 package was used to represent the cells on a pseudo-time trajectory; S4-3: Use the order_cells() function in the Monocle3 package to assign a pseudo time value to each cell along the trajectory from naive B cells to germinal center B cells. Finally, apply the slingshot() function in the Slingshot toolkit to fit a smooth trajectory, which is the TLS maturation trajectory.

5. The model according to claim 2, It is characterized in that In step S5, the regression equation is constructed by: S5-1: The TLS maturation trajectory is evenly divided into several bins. In each bin, the average expression of all cells is taken to calculate the gene expression profile and create a pseudo batch of cells; S5-2: Then, the gene expression profiles associated with mature TLS were extracted to represent the reference cell state at each pseudo time point in the pseudo-batch data and referred to as "reference cells"; S5-3: Calculate the Spearman rank correlation between each B cell to be identified from the immature TLS and each reference cell, and align them to the most appropriate reference cell based on the maximum value of the correlation between the B cell to be identified and the reference cell; S5-4: After comparing all the B cells to be identified with the reference cells, the least squares method was used to fit the regression line based on the pseudo time points and correlation scores. The slope of the regression line represented the overall trend along the TLS maturation trajectory. Based on the linear regression results, the Mfuzz toolkit was used to divide the immature TLS into two different groups. According to the positive or negative slope of the linear regression results, the TLS with a positive slope was classified as TLS tending towards maturity, and the TLS with a negative slope was classified as TLS deviating from maturity. The B cells to be identified were naive B cells or germinal center B cells.

6. The model according to claim 2, It is characterized in that The TLS trending towards maturity and the TLS deviating from maturity are randomly split into training and validation sets in a ratio of 8:2.

Images