Prediction model and construction method for predicting the prognosis of lung cancer after neoadjuvant immunotherapy based on primary tumor

Abstract

The present invention belongs to the field of biological detection technology, and specifically relates to a prediction model for predicting the prognosis of non-small cell lung cancer (NSCLC) after neoadjuvant immunotherapy through primary tumor (PT) and a construction method thereof. The present invention uses a variety of spatial high-throughput sequencing bioinformatics technologies to analyze the key immune cell subsets, immune microenvironment topological characteristic structures and tertiary lymphoid body-related functional cell subsets of primary lesions after neoadjuvant immunotherapy. The proposed prediction model for predicting the prognosis of NSCLC after neoadjuvant immunotherapy through primary lesions and its construction method are more in line with the changes in spatial histological characteristics of the immune microenvironment after treatment. Compared with the previous transcriptome sequencing method without spatial information, it more accurately characterizes the changes in the spatial tumor immune microenvironment after immunotherapy, and can more accurately predict the prognosis of NSCLC after neoadjuvant immunotherapy through cell targets that truly affect the efficacy of immunotherapy.

Description

Technical Field

[0001] The present invention belongs to the technical field of biological detection, and specifically relates to a prediction model for predicting the prognosis after neoadjuvant immunotherapy for non-small cell lung cancer (NSCLC) through primary tumor (PT) and a construction method thereof. Background Art

[0002] Non-small cell lung cancer (NSCLC) is the most common subtype of lung cancer, and its incidence is gradually increasing. Recently, a series of clinical trials of neoadjuvant immunotherapy have shown that patients with resectable non-small cell lung cancer showed significantly increased tumor regression rates and significantly reduced recurrence rates after neoadjuvant immunotherapy combined with chemotherapy, which has led to a gradual increase in the application of neoadjuvant immunotherapy in clinical practice and has become the standard treatment for some NSCLC. However, the efficacy of neoadjuvant immunotherapy varies among different patients, and which immune cell subsets or immune microenvironmental characteristic structures play a key role in affecting the efficacy of neoadjuvant immunotherapy has not been fully studied.

[0003] At present, the few studies exploring the characteristics of the immune microenvironment after neoadjuvant immunotherapy in NSCLC are limited to technologies such as transcriptome sequencing or single-cell transcriptome sequencing. Although data on changes in cell frequency and function have been obtained, there is a lack of characterization of spatial immune microenvironment information. The interaction between various immune cells and tumor cells in the immune microenvironment depends largely on the distance of their spatial position and the formation of microscopic topological structures in the immune microenvironment. For example, even if transcriptome sequencing shows that a certain T cell increases after immunotherapy, if the cell does not infiltrate into the tumor lesion in space and "fight" with tumor cells, whether the many functions of the cell can be exerted needs to be questioned. Therefore, the characterization of the spatial topological characteristics of the immune microenvironment based on spatial proteomics is the right direction to answer the changes in the immune microenvironment after immunotherapy.

[0004] Imaging mass cytometry (IMC) is an emerging technology in spatial proteomics, which couples different target antibodies with rare earth metals and incubates them on the same tissue slide. The sample is scanned point by point using a high-resolution laser system and then analyzed using a mass spectrometer. This technology can detect dozens of protein markers at the same time and preserve the spatial micro information of cells, thereby achieving single-cell phenotypic analysis with spatial resolution.

[0005] The inventors have explored the key immune cell subsets, topological characteristic structures of the immune microenvironment, and functional cell subsets associated with tertiary lymphoid bodies of PT after neoadjuvant immunotherapy by performing imaging mass spectrometry (IMC), single-cell transcriptome sequencing, and spatial transcriptome sequencing on the primary lesions of NSCLC patients who received neoadjuvant immunotherapy (PD-1 monoclonal antibody) combined with chemotherapy. On this basis, a new prediction model for the prognosis of NSCLC after neoadjuvant immunotherapy based on the PT immune microenvironment and its construction method have been developed.

[0006] There is no report on relevant research results in the prior art. Summary of the invention

[0007] In order to overcome the defects and shortcomings in the prior art, the present invention provides a novel prediction model for predicting the prognosis after neoadjuvant immunotherapy for non-small cell lung cancer (NSCLC) through primary tumor (PT) and a construction method thereof.

[0008] In order to achieve the purpose of the present invention, the present invention adopts the following technical scheme:

[0009] In a first aspect, the present invention provides a method for constructing a prediction model for predicting the prognosis of NSCLC after neoadjuvant immunotherapy by PT, wherein the prediction model uses imaging mass spectrometry (IMC) technology, single-cell transcriptome sequencing technology, and spatial transcriptome sequencing technology to predict the prognosis of NSCLC after neoadjuvant immunotherapy, and the prediction model construction method comprises the following steps:

[0010] Step 1: Use IMC technology to detect PT samples to obtain the cell cluster types related to PT and immunotherapy efficacy, so as to determine the differential characteristics of tumor microenvironment components presented by PT after neoadjuvant immunotherapy;

[0011] Step 2: Further analyze the cell neighborhood (CN) and cell interactions of PTs to determine the differences in CN enrichment of PTs with differential immunotherapy responses, so as to determine the differential characteristics of CN components of the tumor microenvironment presented by PTs after neoadjuvant immunotherapy;

[0012] Step 3: Further analyze the spatial location characteristics of CN, define the area dominated by B cells as a "patch", and use this to demarcate the structure of the tertiary lymphoid corpuscle (TLS). Define the area around the "patch" as a complement, and find that the spatial location characteristics are closely related to the key immune microenvironment structure TLS that affects the immune response. By analyzing the differential distribution of CN related to the efficacy inside and outside TLS, it is determined that the key CN that determines the efficacy of immunotherapy is located inside TLS, and the predictive value of CN in TLS for patient prognosis is verified;

[0013] Step 4: Use single-cell transcriptome sequencing and spatial transcriptome sequencing to identify cells within TLS that are associated with the efficacy of neoadjuvant immunotherapy for NSCLC;

[0014] Step 5: Based on the differential characteristics of tumor microenvironment CT components between patients who respond to neoadjuvant immunotherapy and patients who resist treatment obtained in step 1, the differential characteristics of key CNs that determine the immune efficacy located in TLS obtained in step 2, and the differential characteristics of key cell subsets related to the neoadjuvant immune efficacy in TLS obtained in steps 3 and 4, the prediction model is constructed using the least absolute shrinkage and selection operator (LASSO) regression model.

[0015] As an optional approach, in the above-mentioned prediction model construction method, in step 1, the marker set is identified using IMC technology, CT is defined by unsupervised clustering, and a factorial correlation analysis is performed between the CT distribution status and the PT response status to neoadjuvant immunotherapy to determine the differential characteristics of tumor microenvironment components with statistical differences.

[0016] As an optional approach, in the above-mentioned prediction model construction method, in step 2, the identification of CN is based on the central cell and its 10 nearest neighbor cells, and according to the abundance of the original cell clusters in each CN, CNs with different functions are identified, showing differences in the CN enrichment levels of PTs with differential immunotherapy responses, and statistical analysis is performed using a paired Wilcox test.

[0017] As an optional method, in the above prediction model construction method, in step 3, imcRtools (version 1.5.4) was used to determine the "patch" according to the standard of the maximum distance between cells being 20 µm, requiring at least 10 cells to gather. Subsequently, the "patch" was enlarged by 20 µm to fully divide its environment, that is, the "complement", and by comparing the differential distribution of CN inside and outside the TLS, it was determined that the key CN that determines the immune efficacy is located within the TLS, and the Kaplan-Meier curve fitting was used to verify the accuracy of CN in the TLS in predicting prognosis.

[0018] As an optional method, in the above-mentioned prediction model construction method, in step 4, based on unsupervised clustering and spot features, the immune microenvironment is divided into 13 cell groups, namely Macrophage, T cell, B cell, NK cell, Epithelial cell, Neutrophil, Fibroblast, Endothelial cell, Plasma cell, LAMP3+DC, Mast cell, Smoothmuscle cell, and pDC. Subcluster dimensionality reduction and subcluster expression difference molecular interaction and localization analysis are mainly performed on Macrophage, T cell, and B cell, and the TCGA non-small cell lung cancer database and immunotherapy database are used to verify the relationship between the expression of various cell population subtypes and the prognosis after treatment.

[0019] As an optional method, in the above prediction model construction method, in step 5, a treatment-responsive patient is a patient whose comprehensive pathological evaluation shows that the proportion of residual tumor cells in the primary lesion and lymph nodes of lung cancer does not exceed 10%, and who achieves major pathological response (MPR), and is a treatment-responsive patient; a treatment-resistant patient is a patient whose comprehensive pathological evaluation shows that the proportion of residual tumor cells in the primary lesion and lymph nodes of lung cancer exceeds 10%, and who does not achieve major pathological response (MPR), and is a treatment-resistant patient.

[0020] As an optional method, in the above prediction model construction method, in step 5, the LASSO regression model is used to screen variables of CT (including CT with good and poor efficacy), CN in TLS, and FLOR2 cell subpopulations identified by the single-cell idle algorithm, assign correlation coefficients, and construct a model.

[0021] In a second aspect, the present invention provides a prediction model system for predicting the prognosis after neoadjuvant immunotherapy for NSCLC through PT, which is constructed by the prediction model construction method described in the first aspect.

[0022] As an optional method, in the above-mentioned predictive model system, the cell cluster types related to neoadjuvant immunotherapy of PT obtained by IMC technology include ProliferatingT, Epithelial, NK&Epithelial, PD1+CD4T, Neutrophil, NK, Endothelial, MatureB, B&FDC, TRM&CD8T, SmoothMuscle&InfMP, B, TRM.

[0023] Furthermore, the cell cluster types that were negatively correlated with the efficacy of neoadjuvant immunotherapy for PT were ProliferatingT, Epithelial, NK&Epithelial, PD1+CD4T, Neutrophil, and NK.

[0024] Furthermore, the cell cluster types positively correlated with the efficacy of neoadjuvant immunotherapy for PT were Endothelial, MatureB, B&FDC, TRM&CD8T, SmoothMuscle&InfMP, B and TRM.

[0025] As an optional approach, in the above-mentioned predictive model system, CN1:B&TRM, CN6:TRM enriched within TLS significantly enhanced the efficacy of neoadjuvant immunotherapy for PT.

[0026] As an optional method, in the above-mentioned prediction model system, the cells associated with the prognosis of neoadjuvant immunotherapy for NSCLC determined by single-cell transcriptome sequencing technology include HSPA1, CTSK+Mps, IDO1+Mps, MKi 67+Mps, FOLR2+Mps, Exhausted CD4 Tem, Exhausted CD8 Tem, IGHD+B, CRIP1+B, and HSPA1A+B.

[0027] As an optional method, in the above prediction model system, based on the difference characteristics of tumor microenvironment CT components between patients who respond to treatment after neoadjuvant immunotherapy and patients who resist treatment, the difference characteristics of key CNs that determine the efficacy of immunotherapy located in TLS, and the difference characteristics of key cell subsets related to the efficacy of neoadjuvant immunotherapy in TLS, the least absolute shrinkage and selection operator (LASSO) regression model was used to construct the following model: M=0.18 CN1+0.39 CN6+0.91 FOL2 + Mps-0.20 ProliferatingT (CT) -0.35 Neutrophil (CT) was tested and its predictive performance was good.

[0028] Compared with the prior art, the present invention has the following beneficial effects:

[0029] The present invention uses a variety of spatial high-throughput sequencing bioinformatics technologies to analyze key immune cell subsets, immune microenvironment topological characteristic structures and tertiary lymphoid body-related functional cell subsets after neoadjuvant immunotherapy, and proposes a new prediction model and construction method for predicting the prognosis of neoadjuvant immunotherapy for NSCLC through PT. Therefore, the efficacy prediction model of the present invention is more consistent with the changes in the spatial histological characteristics of the immune microenvironment after treatment. Compared with the previous transcriptome sequencing without spatial information, it more accurately characterizes the changes in the spatial tumor immune microenvironment after immunotherapy, and can more accurately predict the prognosis of NSCLC after neoadjuvant immunotherapy through cell targets that truly affect the immune efficacy.

[0030] It should be understood that within the scope of the present invention, the above-mentioned technical features of the present invention and the technical features specifically described below (such as embodiments) can be combined with each other to form a new or preferred technical solution. Due to space limitations, they will not be described one by one here. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 :Flowchart of the method for constructing a prediction model for predicting the prognosis of NSCLC after neoadjuvant immunotherapy through PT. Imaging mass spectrometry (IMC) detection was performed on the primary tumor (PT) of all enrolled patients; single-cell transcriptome sequencing (scRNA sequencing) was performed on 7 enrolled patients, and 9 single-cell transcriptome samples were obtained, including 7 primary tumor tissues and 2 adjacent normal tissues. Spatial transcriptome sequencing was performed on 4 patients, and all of them had matching single-cell transcriptome sequencing.

[0032] Figure 2 : Cell cluster (CT) heatmap. The heatmap shows the maximum-minimum normalized mean marker expression of the 22 CTs defined by unsupervised clustering.

[0033] Figure 3 : Heat map of T cell subclusters. T cells were further clustered into 9 subclusters.

[0034] Figure 4 :Correlation analysis of PT immune efficacy. Revealing the cell cluster subtypes closely related to PT immune efficacy.

[0035] Figure 5 : Representative IMC staining images and comparisons of CT of interest.

[0036] Figure 6 : Cell neighborhood (CN) heatmap. 15 CNs were defined by unsupervised clustering based on the abundance of primitive cell clusters in each CN.

[0037] Figure 7:Topological representation and comparison of CNs of interest. CN1: B&TRM and CN6: TRMenriched in responsive patients in PT were significantly higher than those in resistant patients, with TRM being more critical.

[0038] Figure 8 :Schematic diagram of the Patch algorithm. The area dominated by B cells is defined as a "patch", which is used to demarcate the TLS structure, and the area around the patch is defined as the complement.

[0039] Fig. 9 :The relationship between the distribution of CN inside and outside the patch and the difference in efficacy. CN1: B&TRM is mainly distributed inside the patch, enhancing the immune efficacy; CN6: TRM enriched is slightly more outside the patch than inside the patch, but only CN6 inside the patch enhances the immune efficacy of T; The distribution of CN8 / CN9 / CN11 inside and outside does not affect the efficacy of PT.

[0040] Fig.10 :The predictive value of CN within TLS for prognosis.

[0041] Fig.11 :UMAP diagram of major cell categories. Based on unsupervised clustering and spot features, single-cell data are used to divide the microenvironment into 13 cell groups, including epithelial cells, Macrophage (Mps), T cells, B cells, NK cells, Neutrophil, and Fibroblast.

[0042] Fig.12 :UMAP diagram of cell subpopulations. Mainly for subcluster dimensionality reduction of Macrophage, T cell, and B cell.

[0043] Fig.13 :Heat map of cell subset characteristic gene expression. Mainly displays the standardized average expression level of characteristic genes of Macrophage, T cell, and B.

[0044] Fig.14 :Single cell spatial mapping and TLS annotation. The cell subsets obtained from the single cell data were mapped to their spatial positions, and the TLS positions were defined accordingly. The number of TLS formed by B cell aggregation in responders (N=6) was significantly higher than that in resisters (N=3).

[0045] Fig.15 :Comparison of the number of cell subsets within TLS. The CD8Tem, CRIP1+ B cells, CD4Tem, and FOLR2+ TRM within the TLS of responders are more abundant than those of resisters.

[0046] Fig.16 : Characteristic gene expression of different subgroups of Macrophage.

[0047] Fig.17 :Pseudo-time analysis diagram of Macrophage. FOLR2+ TRM is mainly at the end of cell differentiation.

[0048] Fig.18 : Comparison of FOLR2 expression from RNA sequencing data. The expression of FOLR2 in the response group was significantly higher than that in the resistance group.

[0049] Fig.19 : Based on the difference characteristics of tumor microenvironment CT between patients who respond to PT treatment and those who resist, the difference characteristics of key CN within TLS, and the difference characteristics of key cell subpopulations within TLS, a Lasso regression model and model prediction efficiency evaluation were constructed.

[0050] Fig. 20 :The predictive value of different Macrophage subtype expression levels for prognosis. The prognostic effect of each subtype was verified in the TCGA non-small cell lung cancer database, among which high expression of CTSK+Mps, IDO1+Mps, and MKi 67+Mps was a poor prognostic factor.

[0051] Fig.21 :The predictive value of FOLR2 expression level for prognosis. The TCGA non-small cell lung cancer dataset and the melanoma immunotherapy cohort (PHS000452) and other external datasets verified that high FOLR2 expression is a protective factor for prognosis, and it has a certain predictive value for patient prognosis. DETAILED DESCRIPTION

[0052] The present invention will be further described with reference to specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the scope of the present invention.

[0053] If no specific techniques or conditions are specified in the examples, the techniques or conditions described in the literature in the field or the product instructions are used. If no manufacturer is specified for the reagents or instruments used, they are all conventional products that can be purchased through regular channels.

[0054] The experimental methods in the following examples are conventional methods unless otherwise specified. The experimental materials used in the following examples are commercially available products unless otherwise specified.

[0055] Example:

[0056] Eighteen patients with non-small cell lung cancer and lymph node metastasis were enrolled and received neoadjuvant immunotherapy combined with chemotherapy. The immunotherapy they received included PD-1 monoclonal antibodies such as tislelizumab, pembrolizumab, and sintilimab. This study has been approved by the Ethics Committee of Peking University People's Hospital (approval number: 2022PHB140-001).

[0057] Imaging mass spectrometry (IMC) detection was performed on the primary tumor (PT) of all enrolled patients; single-cell transcriptome sequencing (scRNA sequencing) was performed on 7 enrolled patients, and 9 single-cell transcriptome samples were obtained, including 7 primary tumor tissues and 2 adjacent normal tissues. Spatial transcriptome sequencing was performed on 4 patients, and all of them had supporting single-cell transcriptome sequencing (see Figure 1 ).

[0058] Based on the identification results of 33 markers in IMC (see Table 1), these markers were selected based on their coverage of the lineage markers of major immune cells in the tumor immune microenvironment and functional indicators such as a variety of more important immune checkpoint inhibitor molecules. 22 cell clusters (CT, see Figure 2 ), the heat map shows the maximum-minimum normalized mean marker expression of the 22 clusters, and further clusters T cells into 9 subclusters (see Figure 3 Based on the IMC results, a factorial correlation analysis was performed between the CT distribution and the PT response to neoadjuvant immunotherapy, and the results with statistical significance are shown in Table 2.

[0059] Table 1: Imaging mass cytometry antibody panel

[0060]

[0061]

[0062] Table 2: Factorial analysis of IMC cell clusters and PT response to neoadjuvant immunotherapy

[0063]

[0064] Among them, B cells and TRM were positively correlated with the response of PT to neoadjuvant immune therapy, while NK, Neutrophil, Proliferating T and Epithelial were positively correlated with the resistance to neoadjuvant immune therapy of PT (see Figure 4 ), which shows that PTs show different characteristics of tumor microenvironment components after neoadjuvant immunotherapy, suggesting the direction of further exploring the mechanism of differential response. The representative IMC staining images of these CTs are shown as follows Figure 5 shown.

[0065] To explore the potential mechanisms by which these specific cellular interactions contribute to treatment response, we performed cell neighborhood (CN) and cell interaction analysis using high-resolution spatial data. CNs were identified based on the central cell and its 10 nearest neighbors, and 15 different CNs were identified based on the abundance of the original cell clusters in each CN (see Figure 6 ).

[0066] The difference in CN enrichment of PTs with differential immunotherapy response was shown by combining box plots with dot plots (CN1: B&TRM, CN6: TRM enriched, CN8: B enriched, CN9: Tertiary lymphoid structure compartment, CN11: B&T interaction), and statistical analysis was performed using paired Wilcox test. CN1: B&TRM and CN6: TRM enriched in PTs with response were significantly higher than those in resistant patients (see Figure 7 More importantly, in the actual IMC images, it was found that several key CNs obtained from the differential efficacy analysis were able to focus on the structure closely related to the efficacy - the tertiary lymphoid structure (TLS).

[0067] Subsequently, in order to clarify how these cellular functional modules CN function inside and outside the TLS, the area dominated by B cells was defined as a "patch" to mark the TLS structure, and the area around the patch was defined as the complement (see Figure 8 Using imcRtools (version 1.5.4), “patches” were defined based on a criterion of a maximum intercellular distance of 20 µm, requiring at least 10 cells to be clustered. Subsequently, the “patches” were enlarged by 20 µm to fully partition their environment, i.e., “complementation”.

[0068] The relationship between the distribution of CNs closely related to efficacy inside and outside the patch and the difference in efficacy was analyzed. It was found that for PT, CN1: B&TRM was mainly distributed inside the patch, enhancing the immune efficacy; CN6: TRM enriched was slightly more outside the patch than inside the patch, but only CN6 inside the patch enhanced the immune efficacy of T; the distribution of CN8 / CN9 / CN11 inside and outside the patch did not affect the efficacy of PT (see Fig. 9 ), and further verified the accuracy of CN in TLS in predicting prognosis by using Kaplan-Meier curve fitting (see Fig.10 ).

[0069] The IMC panel of the present invention is limited to 33 markers. In order to analyze the IMC-enriched cell clusters with higher resolution, single-cell transcriptome combined with spatial transcriptome sequencing was performed. Based on unsupervised clustering and spot features, the microenvironment was divided into 13 cell groups, including epithelial cells, macrophage (Mps), T cells, B cells, NK cells, neutrophil, and fibroblasts (see Fig.11 ), and mainly targets Macrophage, T cell, B cell, (see Fig.12 ) performed subcluster dimensionality reduction and subcluster characteristic gene expression analysis (see Fig.13 ). We further mapped the cell subsets obtained from the single-cell data to their spatial locations, and based on this, we defined the locations of TLS formed by B cell aggregation. We found that the number of TLS formed by B cell aggregation in responders (N=6) was significantly greater than that in resisters (N=3) (see Fig.14 ).

[0070] The distribution of different cell subsets within TLS was statistically analyzed and compared. It was found that compared with the resisters, the TLS of responders had more abundant FOLR2+ TRM, CRIP1+ B cells, CD8Tem, and CD4Tem. Combined with the above findings, macrophages represented by FOLR2+TRM within TLS may play a key role (see Fig.15 ).

[0071] Further analysis of the characteristic gene expression of macrophage subsets revealed that FOLR2+ TRM has a strong antigen presentation function (see Fig.16 ). Pseudo-time analysis of macrophage subsets suggests that FOLR2+ TRM is mainly at the terminal end of cell differentiation (see Fig.17 ). In the RNA sequencing data of the included patients, it was found that the expression of FOLR2 in the response group was significantly higher than that in the resistance group (see Fig.18 ).

[0072] The differential characteristics of tumor microenvironment CT components between patients who respond to neoadjuvant immunotherapy and those who resist treatment, the differential characteristics of key CNs that determine the efficacy of immunotherapy located in TLS, and the differential characteristics of key cell subsets related to the efficacy of neoadjuvant immunotherapy in TLS were included in the analysis. The LASSO regression model was used to select positive and negative factors that have an impact on prognosis, and the prediction model was constructed: M=0.18 CN1+0.39 CN6+0.91 FOL2 + Mps-0.20 ProliferatingT (CT) -0.35 Neutrophil (CT), and the AUC curve was used to evaluate the predictive performance of the model (see Fig.19 ).

[0073] The prognostic effect of each Mps subtype was verified using the TCGA non-small cell lung cancer database, among which high expression of CTSK+Mps, IDO1+Mps, and MKi 67+Mps was a poor prognostic factor (see Fig. 20 ). Furthermore, in external datasets such as the TCGA non-small cell lung cancer dataset and the melanoma immunotherapy cohort (PHS000452), it was verified that high expression of FOLR2 is a protective factor for prognosis, and it has a certain predictive value for patient prognosis (see Fig.21 ).

[0074] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A method for constructing a prediction model for predicting the prognosis of non-small cell lung cancer (NSCLC) after neoadjuvant immunotherapy based on primary tumor PT for non-diagnostic purposes, characterized by: The prediction model uses imaging mass spectrometry flow IMC technology, single-cell transcriptome sequencing technology and spatial transcriptome sequencing technology to identify the differential characteristics of key cells in cell clusters CT, cell neighborhood CN, and tertiary lymphoid bodies TLS related to immunotherapy efficacy to predict the prognosis of NSCLC after neoadjuvant immunotherapy. The prediction model construction method includes the following steps: Step 1: Use IMC technology to detect PT samples from patients who respond to neoadjuvant therapy and those who resist therapy to obtain the CT types of cell clusters related to PT and immunotherapy efficacy, so as to determine the differential characteristics of CT components of the tumor microenvironment that determine the efficacy of PT after neoadjuvant immunotherapy; Step 2: Further analyze the cell neighborhood CN and cell interactions of PTs to determine the differences in CN enrichment of PTs with differential immunotherapy responses, so as to determine the differential characteristics of CN components of the tumor microenvironment presented by PTs after neoadjuvant immunotherapy; Step 3: Further analyze the spatial location characteristics of CN, define the area dominated by B cells as a "patch", and use this to demarcate the structure of the tertiary lymphoid body TLS. Define the area around the "patch" as a complement. It is found that the spatial location characteristics are closely related to the key immune microenvironment structure TLS that affects the immune response. By analyzing the differential distribution of CN related to the efficacy inside and outside TLS, it is determined that the key CN that determines the immune efficacy is located inside TLS, and the predictive value of CN in TLS for patient prognosis is verified; Step 4: Use single-cell transcriptome sequencing and spatial transcriptome sequencing to identify cells within TLS that are associated with the efficacy of neoadjuvant immunotherapy for NSCLC; Step 5: Based on the differential characteristics of tumor microenvironment CT components between patients who respond to neoadjuvant immunotherapy and patients who resist treatment obtained in step 1, the differential characteristics of key CNs that determine the immune efficacy located in TLS obtained in step 2, and the differential characteristics of key cell subpopulations related to the neoadjuvant immune efficacy in TLS obtained in steps 3 and 4, the prediction model is constructed using the minimum absolute shrinkage and selection operator LASSO regression model.

2. The prediction model construction method according to claim 1, characterized in that: In step 1, the marker set was identified using IMC technology, CT was defined by unsupervised clustering, and a factorial correlation analysis was performed between the CT distribution and the PT response status to neoadjuvant immunotherapy to determine the differential characteristics of the tumor microenvironment components with statistical differences.

3. The prediction model construction method according to claim 1, characterized in that: In step 2, CNs were identified based on the central cell and its 10 nearest neighbor cells. CNs with different functions were identified based on the abundance of the original cell clusters in each CN. Differences in CN enrichment in PTs that showed differential immunotherapy responses were statistically analyzed using paired Wilcox test.

4. The prediction model construction method according to claim 1, characterized in that: In step 3, the "patch" was determined based on the criterion of a maximum intercellular distance of 20 µm, requiring the aggregation of at least 10 cells. Subsequently, the "patch" was enlarged by 20 µm to fully divide its environment, namely the "complement", and by comparing the differential distribution of CNs inside and outside the TLS, it was determined that the key CNs that determine the efficacy of the immune response are located within the TLS. The Kaplan-Meier curve fitting was used to verify the accuracy of CNs within the TLS in predicting prognosis.

5. The prediction model construction method according to claim 1, characterized in that: In step 4, based on unsupervised clustering and spot features, the immune microenvironment was divided into 13 cell populations, including Macrophage, Tcell, Bcell, NK cell, Epithelialcell, Neutrophil, Fibroblast, Endothelialcell, Plasmacell, LAMP3+DC, Mastcell, Smoothmusclecell, and pDC. Subcluster dimensionality reduction and subcluster expression difference molecular interaction and localization analysis were performed mainly for Macrophage, Tcell, and Bcell. The TCGA non-small cell lung cancer database and immunotherapy database were used to verify the relationship between the expression of various cell population subtypes and the prognosis after treatment.

6. A prediction model system for prognosis after neoadjuvant immunotherapy for NSCLC using PT constructed using the prediction model construction method according to any one of claims 1 to 5.

7. The prediction model system according to claim 6, characterized in that: The cell cluster types related to the efficacy of neoadjuvant immunotherapy for PT obtained using IMC technology include ProliferatingT, Epithelial, NK&Epithelial, PD1+CD4T, Neutrophil, NK, Endothelial, MatureB, B&FDC, TRM&CD8T, SmoothMuscle&InfMP, B, and TRM.

8. The prediction model system according to claim 6, characterized in that: CN1:B&TRM, CN6:TRMenriched within TLS enhance the efficacy of neoadjuvant immunotherapy for PT.

9. The prediction model system according to claim 6, characterized in that: The cells associated with the efficacy of neoadjuvant immunotherapy for NSCLC identified by single-cell transcriptome sequencing technology include HSPA1, CTSK+Mps, IDO1+Mps, MKi 67+Mps, FOLR2+Mps, ExhaustedCD4 Tem, ExhaustedCD8 Tem, IGHD+B, CRIP1+B, and HSPA1A+B.

Images