Constructing a Predictive Model for PD-1 Blockade Therapy in Pan-Cancer Based on Machine Learning

Constructing a Predictive Model for PD-1 Blockade Therapy in Pan-Cancer Based on Machine Learning | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Constructing a Predictive Model for PD-1 Blockade Therapy in Pan-Cancer Based on Machine Learning Hening Li, Shanhang Li, Liang Xiong, Mingxiu Yang, Wei Dai, Xiaoting Luo, and 7 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6317412/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Programmed cell death protein-1 (PD-1) blockade therapy have shown significant efficacy in cancer immunotherapy. However, low response rates and individual variability remain challenges. Currently, a universal biomarker to assess immunotherapy efficacy across various cancer types is lacking. In this study, single-cell RNA sequencing was applied to samples from seven cancer types, alongside bulk RNA-seq data from eight additional cancer types. LASSO regression and 15 machine learning algorithms were employed to construct 152 predictive models for immunotherapy efficacy. The results indicated that CD8 + effector T cells (CD8_Teff) in responders exhibited high infiltration and an activated, exhaustion-like phenotype. A predictive model based on seven effector T cell immunotherapy response genes (ETIRGS) effectively distinguished between responders and non-responders. The high-predicted scoring group exhibited significantly higher infiltration of CD8 + T cells and M1 macrophages than the low-predicted scoring group, along with elevated stromal and immune scores. Macrophages in responders acquired a pro-inflammatory phenotype upon activation by CD8_Teff cells, thereby enhancing the immune response. This study provides potential cross-cancer predictive biomarkers for PD-1 blockade therapy. Health sciences/Biomarkers Health sciences/Oncology CD8 + T effector cells Immune checkpoint blockade Machine learning Pan cancer PD-1 Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Introduction Immune checkpoint blockade (ICB) has brought a historic shift in cancer therapy, transforming therapeutic strategies for various solid tumors. 1 Therapeutic inhibitors targeting cytotoxic T-lymphocyte-associated protein 4 (CTLA-4), programmed cell death protein-1 (PD-1), and its ligand PD-L1 have proven effective in reactivating dysfunctional cytotoxic CD8 + T cells, thereby enabling the elimination of cancer cells. 2 , 3 Studies show that elevated PD-L1 expression in tumors improves the clinical outcomes of PD-1 blockade therapy. 4 , 5 However, more than half of patients fail to respond to PD-1 blockade therapy, even when combined with other treatments. 6 , 7 Additionally, the therapeutic efficacy of anti-PD-1 treatment varies widely across different cancer types and among individual patients. 8 Therefore, developing biomarkers to accurately predict clinical outcomes across cancer patients remains an urgent challenge. Cytotoxic CD8 + T cells are key effector cells responsible for the effectiveness of anti-PD-1 therapy, and their infiltration into the tumor microenvironment (TME) closely associated with treatment response. The interaction between PD-1 and PD-L1 reduces the activation of effector T cells, inhibiting anti-tumor immunity. In the immunosuppressive TME, infiltrating cytotoxic T cells typically exhibit exhaustion or functional impairment, making it difficult to control tumor growth. 9 , 10 Blockade of the PD-1/PD-L1 axis restores the cytotoxic potential of effector T cells and induce tumor regression. 3 , 11 The success of PD-1 blockade therapy depends on the presence of functional cytotoxic effector T cells in the TME. These cells exert their effects through cytotoxic molecules like granzyme and perforin. 12 Thus, a deeper understanding of T cell functions in the TME is critical for uncovering the mechanisms underlying immune therapy and identifying predictive biomarkers. In the context of anti-tumor immunity, biomarker development has become essential for predicting therapeutic responses. While commonly used biomarkers such as PD-L1, tumor mutational burden (TMB), circulating tumor DNA (ctDNA), gut microbiota, and dMMR/MSI-H 13–16 have shown some predictive value in clinical practice, they still have limitations. Their effectiveness is often constrained by technical and biological factors. 17 , 18 To enhance the universality and accuracy of biomarkers, there is an urgent need to develop more objective and feasible tools to predicting treatment outcomes. Recently, several studies have leveraged genomic and transcriptomic data from tumor patients 19 , 20 to identify multiple biomarkers that can better predict responses to ICB therapy. Recent advancements in single-cell RNA sequencing (scRNA-seq) have allowed detailed profiling of tumor-infiltrating immune cells, uncovering diverse T cell subpopulations within the TME. 21 , 22 Simultaneously, machine learning (ML), renowned for its exceptional performance in data mining, has been widely employed across various research domains. The integration of scRNA-seq with ML has not only significantly enhanced the efficiency of mining high-dimensional data but also facilitated the precise identification of specific immune cell subpopulations, the screening of potential biomarkers, and the prediction of therapeutic responses. This study integrated scRNA-seq and bulk RNA-seq to analyze the immune landscape linked to treatment response, highlighting a significant correlation between CD8 effector T cells and immune therapy response. Additionally, LASSO regression analysis was used to select genes associated with immune response, and 152 predictive models were constructed using 15 ML algorithms to predict immune therapy response. A novel effector T cell immunotherapy response genes (ETIRGS) was developed through multi-cohort validation, offering a new framework for predicting PD-1 blockade efficacy. This study underscores the importance of CD8 effector T cells in anti-tumor immunity across various cancers and offers a promising predictive biomarker for optimizing immune therapeutic strategies. Results Pan-cancer Single-cell Expression Atlas of Anti-PD-1 Therapy To assess the tumor immune characteristics across various cancer types, we performed scRNA-seq analysis on 293 samples from 159 patients with seven different cancer types, aiming to construct a single-cell atlas for in-depth analysis (Fig. 1 A; Figure S1 A-B). Each cancer type was associated with corresponding treatment response data (Supplementary Table 1). After rigorous quality control filtering, a single-cell expression atlas consisting of 751,178 cells was generated. Using well-defined marker genes, the cells were categorized into eight major cell types: five immune cell populations (T cells, NK cells, B cells, myeloid cells, and mast cells) and three stromal cell populations (endothelial cells, epithelial cells, and fibroblasts) (Fig. 1 B-C; Figure S1 C-D). Furthermore, the relative percentages of each cell type in the TME across different groups were displayed as bar plots. The results showed a significant increase in the proportions of T cells, B cells, and NK cells in the response (R) group (Fig. 1 D). PD-1 (PDCD1) expression was primarily confined to T cells and NK cells, with significantly higher expression in the R group compared to the non-response (NR) group. Conversely, PD-L1 (CD274) was highly expressed in myeloid cells and mast cells (Fig. 1 E). These findings highlight the complex and heterogeneous cellular composition of the TME in pan-cancer immunotherapy and its relationship with treatment response. T-cell Subpopulation Characteristics and Significant Differences in Response Status It is well-established that T cell infiltration within tumors plays a central role in PD-1 blockade therapy. 23 , 24 Using our high-resolution T-cell atlas across pan-cancer, we identified nine major T-cell subpopulations, including naive T cells (Tn), stress response CD8 + T cells (CD8_Tstr), interferon (IFN) response CD8 + T cells (CD8_Tisg), type 17 helper T(Th17), exhausted CD4 + T cells (CD4_Tex), regulatory T cells (CD4_Treg), transitional effector CD8 + T cells (CD8_tTeff), effector CD8 + T cells (CD8_Teff), and natural killer T cells (NKT) (Fig. 2 A, Figure S2 A). Among these, Tn exhibited typical naive signatures with high expression of naive genes (Fig. 2 B-C). CD8_Tstr and CD8_Tisg exhibited molecular phenotypes shaped by environmental factors. Specifically, CD8_Tstr was characterized by high expression of stress-related heat shock genes (such as HSPA1B and HSPA1A), 25 , 26 along with unique stress response gene signatures, while CD8_Tisg exhibited high levels of IFN-stimulated genes and IFN-response signatures (Fig. 2 B-C). Previous studies have shown that Tstr cells can be detected in situ within the TME of various cancer types, and stress-related genes are significantly upregulated in non-responsive tumors following immune checkpoint blockade therapy. 21 CD4_Tex exhibited high expression of exhaustion markers (such as PDCD1, TIGIT and CTLA4) and transcription factors (TFs) (such as TOX and TOX2), along with upregulated exhaustion-related gene signatures and MARK signaling pathways (Fig. 2 B-C). CD4_Treg exhibited classic Treg gene features, including IL2RA, FOXP3, TNFRSF4, and CTLA4. CD8_tTeff was positioned between activation and cytotoxicity, with elevated expression of glycolysis-related genes (Fig. 2 B-D). CD8_Teff cells showed elevated expression of granzyme genes (GZMA, GZMB, GZMK, and NKG7), IFN-γ (IFNG), Fas ligand (FASLG), and chemokines (CCL4 and CCL5), with a strong enrichment of CD8 cytotoxic gene signatures (Fig. 2 B-D), indicating their effector functions. In contrast to CD8_Teff, NKT cells displayed stronger expression of NK-associated genes (KLRD1, FCGR3A, KLRF1), along with high expression of fatty acid metabolism and pro-apoptotic genes (Fig. 2 B-D). Moreover, we observed increased infiltration of CD8_Teff and CD4_Tex in R groups, while infiltration of CD8_Tisg was decreased (Fig. 2 E-F). To better understand the differentiation trajectories of T cells under anti-PD-1 therapy, we performed Monocle3 analysis, which identified three main branches. All branches originated from Tn cells. Path 1 terminated in CD4_Tex and CD4_Treg, while paths 2 and 3 passed through CD8_Tstr, with path 2 ending in CD8_Teff and NKT, and path 3 concluding with CD8_Tisg (Fig. 3 A). These pathways represent distinct cellular fates, supported by the expression dynamics of relevant gene sets inferred along the pseudotime axis (Fig. 3 B). Effector activation and cytotoxic signatures increased along the pseudotime trajectory of CD8_Teff cells, while exhaustion markers declined. Differential gene expression analysis showed significant upregulation of cytotoxic-related genes (GZMA, GZMB, GZMH, GNLY) and chemokines (CCL5, CCL4) along the pseudotime progression in the developmental lineage of CD8_Teff (Fig. 3 C-D). Furthermore, the developmental state analysis of T cells showed that CD8_Teff from responders were positioned later in the lineage, while CD8_Teff from non-responders were at an earlier stage of development. No significant differences were observed in other T cell subpopulations (Fig. 3 E, Figure S2 B). This suggests that CD8_Teff from responders is more fully differentiated than those from non-responders. CD8_Teff cells in the R group displayed higher cytotoxic characteristics and stronger T cell-mediated anti-tumor activity (Fig. 3 F). These results underscore the pivotal role of T-cell subpopulations, particularly CD8_Teff, in the efficacy of anti-PD-1 therapy. Activated and exhausted-like CD8_Teff Are Associated with Response This study the composition of T-cell subpopulations across different cancer types was analyzed in detail (Fig. 4 A). Notably, CD8_Teff comprised a significant proportion of T cells, and its proportion varied notably across different tumor types. For example, the highest proportion of CD8_Teff was observed in renal clear cell carcinoma (RCC), while the lowest was observed in basal cell carcinoma (BCC) (Fig. 4 B). This suggests that CD8_Teff may exhibit varying degrees of immune activation enrichment in different cancer types. We further investigated the distribution of CD8_Teff across various tumor tissues. In most cancer types, a higher proportion of CD8_Teff was found in the R group (Figure S2 C), which further supports the pivotal role of CD8_Teff in regulating anti-tumor immune responses. Interestingly, within the tumor tissues, CD8_Teff in the R group exhibited higher expression of chemokines (such as CXCL13, CXCR6, and CCL4L2), cytotoxic genes (GZMB, GZMA), and exhaustion markers (PDCD1, CTLA4) compared to the NR group (Fig. 4 C). The elevated expression of these genes is strongly linked to immune regulatory pathways, including T-cell activation, antigen processing, and presentation. In contrast, the NR group displayed upregulation of pathways associated with cell division (Fig. 4 D). PD-1 is a key marker of tumor-reactive T cells in human cancers and is considered a potential biomarker for tumor recognition rather than effector dysfunction. 27 , 28 High PD-1 expression in CD8 + T cells, which exhibit both activation and exhaustion-like phenotypes, is critical for immune checkpoint blockade responses due to their potent anti-tumor activity. 28 , 29 Additionally, PD-1 blockade restores cytotoxic function in PD-1 + CD8 + T cells and enhances immunosuppressive activity in PD-1 + Treg cells. 30 In this study, PD-1 expression was elevated in CD8_Teff cells in the R group, while it was lower in CD4_Treg cells (Figure S2 D). Furthermore, the ratio of PD-1 + CD8_Teff to PD-1 + CD4_Treg was significantly higher in the R group than in the NR group (Figure S2 E). This finding indicates that the PD-1 + CD8 + T-cell to PD-1 + CD4_Treg ratio in the TME is closely associated with the efficacy of immune checkpoint therapy. The ratio of PD-1 + CD8 + T cells to PD-1 + Treg cells is considered a potential predictive marker for PD-1 blockade efficacy, potentially surpassing other factors like PD-L1 expression and tumor mutation burden. 30 In conclusion, these results highlight the potential of CD8_Teff as a critical biomarker for the effectiveness of immune therapies. Next, Hotspot 31 was used to to analyze DEGs (log2FC > 0 and p_val_adj < 0.05) in the CD8_Teff dataset, aiming to identify gene sets closely linked to immune therapy. Five gene modules were identified (Fig. 4 E). Figure S2 F shows the top ten genes for each module. All five modules exhibited significant differences between the R and NR groups. Module 1 demonstrated the strongest correlation in the R group, making it the most representative module (Fig. 4 F). Consequently, module 1 were selected for further analysis, totaling 88 genes. Development and Validation of the ETIRGS To further investigate the immune microenvironment in PD-1 blockade therapy, the RNA-seq expression profiles of 172 pan-cancer bulk samples were analyzed. A volcano plot was generated to display DEGs between the R and NR groups, identifying 635 up-regulated and 223 down-regulated genes (log2FC > 1 or < -1, p_val_adj < 0.05) (Fig. 4 G). Given the pivotal role of CD8_Teff in PD-1 blockade therapy, we initially intersected the 88 genes from module 1 with the 635 up-regulated genes, resulting in 16 candidate genes (Fig. 4 H). These genes were further optimized using LASSO regression analysis, identifying 7 ETIRGS: CXCL13, GZMB, KLRD1, PDCD1, LSP1, BATF and ABI3. Based on these 7 genes, we constructed 152 ML prediction models using 15 algorithms. The models were compared by calculating their F-scores, identifying the best model (NN-MLP + Lasso-CV; parameter settings: cutoff: 0.25, lr: 0.001, bs: 115, ep: 50, dropout: 0.5) (Fig. 5 A; Figure S3 A-B). To further assess each gene's contribution to the model, we compared their weights and evaluated their importance using the SHAP framework. The results indicated that CXCL13 and GZMB were identified as top contributors with higher model weights. (Fig. 5 B-C). Correlation analysis of the 7 ETIRGS revealed positive associations, with CXCL13 showing the strongest correlation with GZMB and PDCD1. (Figure S3 C). Notably, the expression of all 7 ETIRGS was significantly higher in the R group than in the NR group (Fig. 5 D), indicating that these genes could serve as independent predictive biomarkers for PD-1 blockade efficacy. Immune cell infiltration analysis using the CIBERSORT algorithm revealed significantly higher levels of CD8 T cells and M1 macrophages in the high-predicted scoring group compared to the low-predicted group (Fig. 5 E). Additionally, stromal, immune, and ESTIMATE scores, evaluated through the ESTIMATE function, were also significantly elevated in the high-predicted group (Fig. 5 F). This further supports the potential of ETIRGS as reliable predictive biomarkers for the efficacy of PD-1 blockade therapy. Impact of Macrophage Phenotypes and CD8_Teff Interaction on Treatment Efficacy Myeloid cells are pivotal in tumor immunity. 32 , 33 In this study, myeloid cells in the pan-cancer dataset were subdivided into 8 distinct clusters (Figure S4A-B). To explore whether interactions between CD8_Teff and myeloid cells contribute to immune therapy response, using the CellChat method, it was found that ligand-receptor interactions between CD8_Teff and CXCL10 + macrophages were significantly enriched in the R group, while the NR group showed stronger interactions between CD8_Teff and SPP1 + macrophages (Fig. 6 A-B). Macrophages can polarize into pro-inflammatory M1 or anti-inflammatory M2 phenotypes, thereby modulating T cell responses in anti-tumor immunity. 34 , 35 Further analysis revealed that CXCL10 + macrophages displayed M1 characteristics, while SPP1 + macrophages were predominantly polarized to the M2 phenotype (Fig. 6 C-D, Figure S4C-D). CXCL10 is a well-known chemokine that mediates T cell recruitment. 36 , 37 The high expression of CXCL10 in CXCL10 + macrophages supports their pro-inflammatory properties, suggesting that the pro-inflammatory macrophage phenotype is likely shaped by active anti-tumor immune responses within the TME. Previous studies have shown that M2-like SPP1 + macrophages are closely linked to angiogenesis and poor prognosis in patients. 38 Subsequently, we specifically focused on ligand-receptor interactions. In the R group, the number of ligand-receptor interactions between CD8_Teff and CXCL10 + macrophages was the highest among the myeloid cell subpopulations. Importantly, CD8_Teff cells produced multiple ligands (CCL5 and CCL4), which have significant regulatory potential. These ligands bind to the receptors (CCR5 and CCR1) on CXCL10 + macrophages, promoting the migration of pro-inflammatory macrophages to the tumor site, thereby cooperating to drive the immune response (Figs. 6 E-F). Overall, in responsive tumors, CD8_Teff cells likely shape the phenotype of pro-inflammatory macrophages. The pro-inflammatory signaling molecules expressed by CD8_Teff and CXCL10 + macrophages may further enhance anti-tumor immunity. In contrast, the M2-type SPP1 + macrophages contribute to an immune-suppressive TME. Discussion PD-1 blockade therapy has shown broad clinical potential across various tumor types. 39 The critical role of tumor-infiltrating T cells in immune responses is well-established. 40 However, a universal biomarker to assess immune therapy responsiveness across cancer types is still lacking. Therefore, there is an urgent need to develop a universal metric to accurately evaluate immune therapy efficacy across cancer types. To address this gap, we integrated gene expression profiles from diverse tumor samples to explore the heterogeneity of tumor-infiltrating T cells across cancers and identify key immune features linked to immune therapy response. Our findings identify that CD8_Teff can serve as a potential predictive biomarker for PD-1 blockade therapy. The differentiation status, infiltration levels, and high expression of effector activation and cytotoxicity genes in CD8_Teff correlate closely with treatment response. Furthermore, we found that CD8_Teff from responders strongly interact with pro-inflammatory macrophages (CXCL10 + macrophages) in the TME, primarily through ligand-receptor pairs such as CCL5-CCR5, CCL5-CCR1, and CCL4-CCR5. These cells are co-enriched in the TME of responders, suggesting that the significant response to PD-1 blockade therapy in tumor patients may be driven by the interaction between CD8 effector T cells and pro-inflammatory macrophages, thereby enhancing anti-tumor immunity. Next, using Lasso regression, we applied 15 ML methods to fit 152 models, ultimately establishing the ETIRGS model, which shows significant potential in predicting immune therapy efficacy. The optimal model was derived using the NN-MLP + Lasso-CV algorithm. Compared to traditional single biomarkers, the ETIRGS model offers superior and more stable performance. As an independent indicator, it effectively complements and overcomes the limitations of traditional biomarkers. These results provide valuable insights for future studies on immune therapy responsiveness. The ETIRGS comprises of seven key genes: CXCL13, GZMB, KLRD1, PDCD1, LSP1, BATF and ABI3. Several of these genes have been identified as potential predictive biomarkers for ICB therapy. Among these, CXCL13 holds substantial academic value for exploration. Originally termed B-cell attracting chemokine 1 (BCA-1) due to its strong chemotactic effect on B cells, 41 , 42 recent studies revealed that CXCL13 is secreted by diverse T cell populations, such as T follicular helper cells (TFH), tumor-infiltrating CD8 + T cells, and exhausted CD8 + T cells. 43 In multiple cancer types, CXCL13 expression closely correlates with ICB response. 44 – 46 For instance, in non-small cell lung cancer (NSCLC), CXCL13 expression was observed in PD-1 high cytotoxic T cells, highlighting its potential as a biomarker for PD-1 blockade therapy response. 27 In triple-negative breast cancer (TNBC) patients, CXCL13 + PD-1 + T cells clonally expand significantly after PD-1 blockade therapy, exhibiting cytotoxic activity and exhaustion markers. 47 Similarly, CXCL13 + T cells were identified in liver metastasis of gallbladder cancer, correlating with higher immune scores. 48 Additionally, in EB virus-related gallbladder cancer, PD-1 blockade therapy activated CXCL13-expressing effector cell subsets. 49 As a further exploration, our study integrated analyses of various cancer samples and confirmed that CD8_Teff cells from responders exhibit both activation and exhaustion characteristics, capable of activating CXCL10 + macrophages into a pro-inflammatory phenotype. Furthermore, CXCL13 expression was positively correlated with inhibitory genes (e.g., PDCD1) and cytotoxic molecules (e.g., GZMB). We propose that CXCL13 may serve as a unifying marker for CD8 effector T cells in the TME, with broad potential to predict responses to PD-1 blockade therapy. The potential role of CXCL13 as a positive predictive biomarker for immune therapy warrants further investigation. Moreover, GZMB, a serine protease, is a key factor mediating apoptosis in target cells by NK cells and cytotoxic CD8 T cells. 50 , 51 GZMB + CD8 T cells are significant predictive markers for ICB therapy response and can effectively predict treatment sensitivity. 52 Compared to previous studies, the combined prediction model of CXCL13 and GZMB for PD-1 blockade therapy efficacy outperforms single biomarker models, potentially helping to identify patients who are most likely to benefit from immune therapy. Our predictive model strongly supports patient stratification based on immune treatment response, facilitating the development of personalized treatment strategies. Nevertheless, our study has certain limitations. Firstly, this study is retrospective, and all data and clinical information are derived from public databases. Larger sample studies are needed for future validation. Secondly, the role of ETIRGS in excluded solid tumors remains unclear, necessitating further research to validate its universality and clinical applicability across diverse cancer types. In conclusion, our study identifies key immune cell subsets associated with the response to PD-1 blockade therapy across various cancers and provides new universal biomarkers for predicting therapeutic efficacy. Further studies are required to validate our findings. Material and methods Data Collection This study included nine scRNA-seq datasets, comprising single-cell transcriptomic data from 258 samples of 159 individuals. These datasets represent seven different cancer types, with clinical information on treatment response status for each cancer type (Fig. 1 A, Supplementary Table 1). Additionally, bulk RNA-seq datasets from eight cancer types treated with anti-PD-1 therapy, comprising 172 samples, were obtained. Only tumor samples obtained post-immunotherapy were retained for analysis. Supplementary Table 2 provides the dataset accession numbers and references. Single-Cell Data Processing ScRNA-seq data integration and analysis were conducted in R (v.4.0.3) using the Seurat (v.4.3) package. The following procedure was adopted: First, the expression matrix of each sample was imported into R with the Read10X() function, and a Seurat object was created, integrating the corresponding clinical information. The Seurat object was then converted into a Scanpy object (Python v.3.9.0) using the SeuratDisk package for further analysis. The Scrublet package (v.0.2.3) was used to predict and filter potential doublets. Quality control was based on the following criteria: (1) total UMI count per cell < 25,000; (2) the number of detected genes between 800 and 5,000; (3) excluding cells with 98% of genes expressed; and (4) the percentage of mitochondrial genes < 10%. Next, the raw counts were normalized and log-transformed using the Scanpy package (v.1.9.8). After filtering for highly variable genes (HVGs), the first 10 principal components (PCs) were selected for subsequent analysis. The BBKNN algorithm was used to construct a batch-balanced k-nearest neighbor (kNN) graph. 53 After batch effect correction, cells were clustered using the Leiden algorithm. Differentially expressed genes (DEGs) were identified using the sc.tl.rank_genes_group() function in Scanpy with the 'use_raw = True' parameter. 54 Based on the DEGs of each cluster and known marker genes, major cell types were identified, including T cells (CD3D, CD3E), NK cells (NKG7, KLRD1, NCAM), B cells (CD79A, MZB1), myeloid cells (LYZ, CD68), mast cells (TPSB1, KIT), endothelial cells (VWF, EMCN), epithelial cells (KRT5, KRT14), and fibroblasts (COL6A2, COL1A2). Clusters expressing multiple standard marker genes were considered as doublet cells. Doublets and undefined cells were excluded, resulting in a total of 630,638 cells for further analysis. The Scanpy object was then converted back into a Seurat object using the sceasy package (v.0.0.7). Subclustering of the main cell types was performed based on the original expression matrix, using the same preprocessing steps, which resulted in the identification of subcluster structures. Identification of T-cell Types and Functional States To determine T-cell types and their functional states, we followed the methodology proposed by Chu et al.. 21 First, DEGs for major cell types were identified using the FindAllMarkers() function from the Seurat package. Subsequently, a bubble plot was generated to visualize the DEGs and T-cell marker genes (Fig. 2 B). Finally, the AddModuleScore() function in Seurat was employed to calculate gene feature scores for selected gene sets linked to functional states within each T-cell cluster. Tissue Distribution of Clusters To quantify the tissue preference of each cluster across different immune response groups, the ratio of observed to expected cell counts (Ro/e) was calculated for each cluster. 55 , 56 Specifically, a chi-square test was performed to obtain the expected cell counts for each cell cluster and tissue combination. If Ro/e > 1, it indicated a higher presence of the cluster in the specific immune response group. Single-Cell Differentiation Trajectory Inference Monocle 3 (v.1.3.7) 57 was used to infer the differentiation trajectories of T cells. The expression matrix of the T cell Seurat object was input into Monocle 3, and a cds object was created using the new_cell_data_set() function. Subsequently, dimensionality reduction, cell clustering, and differentiation trajectory inference were performed. Differential gene analysis along the trajectory was conducted using the graph_test() function. Gene Feature Calculation Twenty-three typical T cell cytotoxicity features were collected based on previous studies on T cells. 21 Genes related to T cell-mediated immune responses against tumor cells were obtained from Li et al.'s study. 58 The gene sets for M1 and M2 macrophage signatures were derived from Bi et al.'s study. 59 The AddModuleScore() function in the Seurat package was used to score cell subpopulations based on gene sets related to T cell cytotoxicity, T cell-mediated immune responses to tumor cells, and macrophage phenotype. Statistical significance of the scores was assessed using the Wilcoxon test. Cell-Cell Interaction Analysis Cell-cell interactions between T cell subpopulations were systematically inferred and visualized using CellChat (v.1.6.1). 60 To explore the impact of PD-1 blockade on intercellular communication network between cell subpopulations, CellChat objects generated using the gene expression matrices from the R and NR groups. The mergeCellChat() function was employed to combine the CellChat objects from both groups, enabling the analysis of differences in the frequency and intensity of inferred intercellular interactions. Subsequently, the netVisual_bubble() function was applied to visualize significant ligand-receptor pair changes in intercellular interactions. Hotspot Gene Module Analysis Hotspot (v.1.1.1) is a tool used to identify co-expressed gene modules in scRNA-seq datasets. 31 To identify gene modules associated with immune therapy responses, the DEGs of CD8_Teff cells from the R and NR groups were first extracted. The create_knn_graph() function from the Hotspot package was then used to construct a k-nearest neighbor (KNN) graph, selecting 30 neighboring genes to generate a gene adjacency matrix. Next, compute_autocorrelations() was employed to calculate the autocorrelation of the genes. Genes with an FDR < 0.001 were selected for subsequent local correlation calculations. Finally, gene module clustering was performed using the create_modules() function to identify gene modules related to immune therapy response. Bulk RNA-seq Analysis Gene expression data and clinical information for the patients were retrieved from the GEO database ( https://www.ncbi.nlm.nih.gov/geo/ ). The raw sequencing data for the samples were aligned and annotated using the GRCh38.p13 reference genome. Raw sequencing data for gastric cancer were downloaded from the European Nucleotide Archive ( https://www.ebi.ac.uk/ena/browser/view/PRJNA557841 ) and subjected to quality control using FastQC, assessing data quality, base distribution, GC content, and adapter contamination. Adapter sequences and low-quality bases were trimmed using the Trim Galore tool. The trimmed reads were then aligned to the GRCh38 reference genome using HISAT2, with the alignment results stored in sorted BAM file format. Gene expression levels were quantified using the featureCounts tool, based on the Gencode v47 genome annotation. The resulting raw count matrix was used for subsequent analyses. ETIRGS Signature Identification Differential expression analysis was performed using the "limma" method, with threshold criteria of logFC > 1 or < -1 and p_val_adj < 0.05. Up-regulated genes were intersected with module genes to identify candidate genes. Subsequently, a rigorous least absolute shrinkage and selection operator (LASSO) regression was conducted using the glmnet (v.4.1.8) package in R to select genes most strongly associated with immune therapy response. Establishment of the ETIRGS Model In the model construction process, a total of 15 ML algorithms were employed, including Neural Network (MLP), Logistic Regression, Linear Discriminant Analysis (LDA), Quadratic Discriminant Analysis (QDA), k-Nearest Neighbor (k-NN), Decision Tree, Random Forest, XGBoost Classification, Ridge Regression, Lasso Regression, Elastic Net Regression, Support Vector Machine (SVM), Gradient Boosting Machine (GBM), Stepwise Logistic Regression, and Naive Bayes Algorithm. In total, 152 ML models were generated using these methods. Each algorithm was trained using a 10-fold cross-validation approach on the training dataset to optimize the model. The performance of the models was then evaluated on the test dataset. The accuracy, recall, and F-score for each model were computed and recorded for each fold of the cross-validation. For training the SVM model, a comparison was made among different kernel functions (linear, radial basis, and polynomial), and the best kernel was selected. To enhance model stability and accuracy, an ensemble learning approach was applied, aggregating predictions from different algorithms through a voting mechanism. Specifically, weights were assigned based on each model's performance on the validation set, and a weighted voting strategy was employed to derive the final prediction result. The prediction results of each algorithm were assigned different weights according to their performance in cross-validation, ultimately generating an integrated prediction result to improve the overall model’s stability and accuracy. CIBERSORT CIBERSORT was applied to assess differences in immune cell infiltration abundance between high and low ETIRGS prediction score groups. The CIBERSORT () function was employed to quantify the scores for each immune cell subgroup. The relative abundance of all immune cell types in each group was predicted using the leukocyte characteristic matrix (LM22). Statistical Analysis Wilcoxon rank-sum test and Kruskal-Wallis test were utilized to compare numerical variables between response groups and across different cell clusters. Statistical significance was defined as p < 0.05 for all analyses. Data analysis and visualization were performed using R (v.4.0.3) and Python (v.3.9.0). Declarations Acknowledgments We are grateful to the contributors to the public databases used in this study and all the authors of the study. Author contributions Jie Ma, Juliang He, Haijun Tang and Yun Liu contributed to the study design and critical revision of the manuscript. Hening Li, Shanhang Li, Liang Xiong and Mingxiu Yang carried out the study and drafted the manuscript. Wei Dai, Xiaoting Luo, Shangyu Liu, Danting Xiao and Binyuan Ning analyzed the data. All authors read and approved the final manuscript. Funding This study was supported by the Natural Science Foundation of Guangxi Zhuang Autonomous Region [Grant No.2025GXNSFAA069103]. Data availability The datasets presented in this study are available in an online repository. The data that support the findings of this study are available from the following databases: GEO database (https://www.ncbi.nlm.nih.gov/geo/), NGDC database (https://ngdc.cncb.ac.cn/omix/), Single Cell Portal (https://singlecell.broadinstitute.org/) and EBI ENA (https://www.ebi.ac.uk/ena/). The accession numbers of the repository can be found in the Supplementary Material. Competing interests The authors declare no competing interests. Ethics approval and informed consent This study utilized human data obtained from publicly available databases. The study protocol was approved by the Institutional Review Board/Ethics Committee of the First Affiliated Hospital of Guangxi Medical University(Numbers: 2024-E0885). Footnotes Hening Li, Shanhang Li, Liang Xiong and Mingxiu Yang contributed equally to this work. Contributor Information Jie Ma: [email protected] , Juliang He: [email protected] , Haijun Tang: [email protected] , Yun Liu: [email protected] . References Kraehenbuehl, L., Weng, C. H., Eghbali, S., Wolchok, J. D. & Merghoub, T. Enhancing immunotherapy in cancer by targeting emerging immunomodulatory pathways. Nat. Rev. Clin. Oncol. 19 , 37–50. 10.1038/s41571-021-00552-7 (2022). Sharma, P. & Allison, J. P. Immune checkpoint targeting in cancer therapy: toward combination strategies with curative potential. Cell 161 , 205–214. 10.1016/j.cell.2015.03.030 (2015). Zou, W., Wolchok, J. D. & Chen, L. PD-L1 (B7-H1) and PD-1 pathway blockade for cancer therapy: Mechanisms, response biomarkers, and combinations. Sci. Transl Med. 8 , 328rv324. 10.1126/scitranslmed.aad7118 (2016). Carbone, D. P. et al. First-Line Nivolumab in Stage IV or Recurrent Non-Small-Cell Lung Cancer. N Engl. J. Med. 376 , 2415–2426. 10.1056/NEJMoa1613493 (2017). Rizvi, N. A. et al. Cancer immunology. Mutational landscape determines sensitivity to PD-1 blockade in non-small cell lung cancer. Science 348 , 124–128. 10.1126/science.aaa1348 (2015). Darvin, P., Toor, S. M., Sasidharan Nair, V. & Elkord, E. Immune checkpoint inhibitors: recent progress and potential biomarkers. Exp. Mol. Med. 50 , 1–11. 10.1038/s12276-018-0191-1 (2018). He, X. & Xu, C. Immune checkpoint signaling and cancer immunotherapy. Cell. Res. 30 , 660–669. 10.1038/s41422-020-0343-4 (2020). Zhao, B., Zhao, H. & Zhao, J. Efficacy of PD-1/PD-L1 blockade monotherapy in clinical trials. Ther. Adv. Med. Oncol. 12 , 1758835920937612. 10.1177/1758835920937612 (2020). Mellman, I., Coukos, G. & Dranoff, G. Cancer immunotherapy comes of age. Nature 480 , 480–489. 10.1038/nature10673 (2011). Zheng, C. et al. Landscape of Infiltrating T Cells in Liver Cancer Revealed by Single-Cell Sequencing. Cell 169, 1342–1356 e1316, (2017). 10.1016/j.cell.2017.05.035 Dong, H. et al. Tumor-associated B7-H1 promotes T-cell apoptosis: a potential mechanism of immune evasion. Nat. Med. 8 , 793–800. 10.1038/nm730 (2002). Russell, J. H. & Ley, T. J. Lymphocyte-mediated cytotoxicity. Annu. Rev. Immunol. 20 , 323–370. 10.1146/annurev.immunol.20.100201.131730 (2002). Wang, Y. et al. FDA-Approved and Emerging Next Generation Predictive Biomarkers for Immune Checkpoint Inhibitors in Cancer Patients. Front. Oncol. 11 , 683419. 10.3389/fonc.2021.683419 (2021). Zhou, C. B., Zhou, Y. L. & Fang, J. Y. Gut Microbiota in Cancer Immune Response and Immunotherapy. Trends Cancer . 7 , 647–660. 10.1016/j.trecan.2021.01.010 (2021). Ahn, S. et al. Changes and prognostic values of tumor-infiltrating lymphocyte subsets after primary systemic therapy in breast cancer. PLoS One . 15 , e0233037. 10.1371/journal.pone.0233037 (2020). Zhang, Q. et al. Prognostic and Predictive Impact of Circulating Tumor DNA in Patients with Advanced Cancers Treated with Immune Checkpoint Blockade. Cancer Discov . 10 , 1842–1853. 10.1158/2159-8290.CD-20-0047 (2020). Doroshow, D. B. et al. PD-L1 as a biomarker of response to immune-checkpoint inhibitors. Nat. Rev. Clin. Oncol. 18 , 345–362. 10.1038/s41571-021-00473-5 (2021). Zouein, J., Kesrouani, C. & Kourie, H. R. PD-L1 expression as a predictive biomarker for immune checkpoint inhibitors: between a dream and a nightmare. Immunotherapy 13 , 1053–1065. 10.2217/imt-2020-0336 (2021). Bai, X. et al. Development and validation of a genomic mutation signature to predict response to PD-1 inhibitors in non-squamous NSCLC: a multicohort study. J. Immunother Cancer . 8 10.1136/jitc-2019-000381 (2020). Chowell, D. et al. Improved prediction of immune checkpoint blockade efficacy across multiple cancer types. Nat. Biotechnol. 40 , 499–506. 10.1038/s41587-021-01070-8 (2022). Chu, Y. et al. Pan-cancer T cell atlas links a cellular stress response state to immunotherapy resistance. Nat. Med. 29 , 1550–1562. 10.1038/s41591-023-02371-y (2023). Zheng, L. et al. Pan-cancer single-cell landscape of tumor-infiltrating T cells. Science 374 , abe6474. 10.1126/science.abe6474 (2021). Yost, K. E. et al. Clonal replacement of tumor-specific T cells following PD-1 blockade. Nat. Med. 25 , 1251–1259. 10.1038/s41591-019-0522-3 (2019). Tay, C., Tanaka, A. & Sakaguchi, S. Tumor-infiltrating regulatory T cells as targets of cancer immunotherapy. Cancer Cell. 41 , 450–465. 10.1016/j.ccell.2023.02.014 (2023). Fang, X., Bogomolovas, J., Trexler, C. & Chen, J. The BAG3-dependent and -independent roles of cardiac small heat shock proteins. JCI Insight . 4 10.1172/jci.insight.126464 (2019). Yang, Y. et al. Pan-cancer single-cell dissection reveals phenotypically distinct B cell subtypes. Cell 187, 4790–4811 e4722, (2024). 10.1016/j.cell.2024.06.038 Thommen, D. S. et al. A transcriptionally and functionally distinct PD-1(+) CD8(+) T cell pool with predictive potential in non-small-cell lung cancer treated with PD-1 blockade. Nat. Med. 24 , 994–1004. 10.1038/s41591-018-0057-z (2018). Mietz, J. et al. Human effector CD8(+) T cells with an activated and exhausted-like phenotype control tumour growth in vivo in a humanized tumour model. EBioMedicine 106 , 105240. 10.1016/j.ebiom.2024.105240 (2024). Oliveira, G. et al. Preexisting tumor-resident T cells with cytotoxic potential associate with response to neoadjuvant anti-PD-1 in head and neck cancer. Sci. Immunol. 8 , eadf4968. 10.1126/sciimmunol.adf4968 (2023). Kumagai, S. et al. The PD-1 expression balance between effector and regulatory T cells predicts the clinical efficacy of PD-1 blockade therapies. Nat. Immunol. 21 , 1346–1358. 10.1038/s41590-020-0769-3 (2020). DeTomaso, D. & Yosef, N. Hotspot identifies informative gene modules across modalities of single-cell genomics. Cell Syst 12, 446–456 e449, (2021). 10.1016/j.cels.2021.04.005 Zhang, L. et al. Single-Cell Analyses Inform Mechanisms of Myeloid-Targeted Therapies in Colon Cancer. Cell 181, 442–459 e429, (2020). 10.1016/j.cell.2020.03.048 Mantovani, A., Allavena, P., Marchesi, F. & Garlanda, C. Macrophages as tools and targets in cancer therapy. Nat. Rev. Drug Discov . 21 , 799–820. 10.1038/s41573-022-00520-5 (2022). Nalio Ramos, R. et al. Tissue-resident FOLR2(+) macrophages associate with CD8(+) T cell infiltration in human breast cancer. Cell 185, 1189–1207 e1125, (2022). 10.1016/j.cell.2022.02.021 Lavin, Y. et al. Innate Immune Landscape in Early Lung Adenocarcinoma by Paired Single-Cell Analyses. Cell 169, 750–765 e717, (2017). 10.1016/j.cell.2017.04.014 House, I. G. et al. Macrophage-Derived CXCL9 and CXCL10 Are Required for Antitumor Immune Responses Following Immune Checkpoint Blockade. Clin. Cancer Res. 26 , 487–504. 10.1158/1078-0432.CCR-19-1868 (2020). Peperzak, V. et al. CD8 + T cells produce the chemokine CXCL10 in response to CD27/CD70 costimulation to promote generation of the CD8 + effector T cell pool. J. Immunol. 191 , 3025–3036. 10.4049/jimmunol.1202222 (2013). Cheng, S. et al. A pan-cancer single-cell transcriptional atlas of tumor infiltrating myeloid cells. Cell 184, 792–809 e723, (2021). 10.1016/j.cell.2021.01.010 Sharma, P. et al. Immune checkpoint therapy-current perspectives and future directions. Cell 186 , 1652–1669. 10.1016/j.cell.2023.03.006 (2023). St Paul, M. & Ohashi, P. S. The Roles of CD8(+) T Cell Subsets in Antitumor Immunity. Trends Cell. Biol. 30 , 695–704. 10.1016/j.tcb.2020.06.003 (2020). Legler, D. F. et al. B cell-attracting chemokine 1, a human CXC chemokine expressed in lymphoid tissues, selectively attracts B lymphocytes via BLR1/CXCR5. J. Exp. Med. 187 , 655–660. 10.1084/jem.187.4.655 (1998). Gunn, M. D. et al. A B-cell-homing chemokine made in lymphoid follicles activates Burkitt's lymphoma receptor-1. Nature 391 , 799–803. 10.1038/35876 (1998). Xu, T. et al. Identification of CXCL13 as a potential biomarker in clear cell renal cell carcinoma via comprehensive bioinformatics analysis. Biomed. Pharmacother . 118 , 109264. 10.1016/j.biopha.2019.109264 (2019). Groeneveld, C. S. et al. Tertiary lymphoid structures marker CXCL13 is associated with better survival for patients with advanced-stage bladder cancer treated with immunotherapy. Eur. J. Cancer . 148 , 181–189. 10.1016/j.ejca.2021.01.036 (2021). Yang, M. et al. CXCL13 shapes immunoactive tumor microenvironment and enhances the efficacy of PD-1 checkpoint blockade in high-grade serous ovarian cancer. J. Immunother Cancer . 9 10.1136/jitc-2020-001136 (2021). Goswami, S. et al. ARID1A mutation plus CXCL13 expression act as combinatorial biomarkers to predict responses to immune checkpoint therapy in mUCC. Sci. Transl Med. 12 10.1126/scitranslmed.abc4220 (2020). Bassez, A. et al. A single-cell map of intratumoral changes during anti-PD1 treatment of patients with breast cancer. Nat. Med. 27 , 820–832. 10.1038/s41591-021-01323-8 (2021). Xie, L., Ning, Z., Hua, Y., Wang, P. & Meng, Z. Single-cell transcriptome analysis revealed the immune profile of PD-1 blockade in gallbladder carcinoma liver metastasis. Hepatol. Commun. 7 10.1097/HC9.0000000000000131 (2023). Qiu, M. Z. et al. Dynamic single-cell mapping unveils Epstein–Barr virus-imprinted T-cell exhaustion and on-treatment response. Signal. Transduct. Target. Ther. 8 , 370. 10.1038/s41392-023-01622-1 (2023). Mueller, A. et al. Transcriptional and Clonal Characterization of Cytotoxic T Cells in Crescentic Glomerulonephritis. J. Am. Soc. Nephrol. 34 , 1003–1018. 10.1681/ASN.0000000000000116 (2023). Wang, Y. et al. Targeting CD96 overcomes PD-1 blockade resistance by enhancing CD8 + TIL function in cervical cancer. J. Immunother Cancer . 10 10.1136/jitc-2021-003667 (2022). Wang, X. Q. et al. Spatial predictors of immunotherapy response in triple-negative breast cancer. Nature 621 , 868–876. 10.1038/s41586-023-06498-3 (2023). Polanski, K. et al. BBKNN: fast batch alignment of single cell transcriptomes. Bioinformatics 36 , 964–965. 10.1093/bioinformatics/btz625 (2020). Traag, V. A., Waltman, L. & van Eck, N. J. From Louvain to Leiden: guaranteeing well-connected communities. Sci. Rep. 9 , 5233. 10.1038/s41598-019-41695-z (2019). Guo, X. et al. Global characterization of T cells in non-small-cell lung cancer by single-cell sequencing. Nat. Med. 24 , 978–985. 10.1038/s41591-018-0045-3 (2018). Zhang, L. et al. Lineage tracking reveals dynamic relationships of T cells in colorectal cancer. Nature 564 , 268–272. 10.1038/s41586-018-0694-x (2018). Cao, J. et al. The single-cell transcriptional landscape of mammalian organogenesis. Nature 566 , 496–502. 10.1038/s41586-019-0969-x (2019). Liu, Z. et al. Progenitor-like exhausted SPRY1(+)CD8(+) T cells potentiate responsiveness to neoadjuvant PD-1 blockade in esophageal squamous cell carcinoma. Cancer Cell. 41 (e1859), 1852–1870. 10.1016/j.ccell.2023.09.011 (2023). Bi, K. et al. Tumor and immune reprogramming during immunotherapy in advanced renal cell carcinoma. Cancer Cell 39, 649–661 e645, (2021). 10.1016/j.ccell.2021.02.015 Jin, S. et al. Inference and analysis of cell-cell communication using CellChat. Nat. Commun. 12 , 1088. 10.1038/s41467-021-21246-9 (2021). Additional Declarations No competing interests reported. Supplementary Files TableS1SampleinformationofscRNAseqdata.xlsx TableS2SampleinformationofbulkRNAseqdata.xlsx FigureS1S4.pdf Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6317412","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":436214969,"identity":"c0ea318f-d920-4acb-9b01-c56a9c3f171b","order_by":0,"name":"Hening Li","email":"","orcid":"","institution":"the First Affiliated Hospital of Guangxi Medical University","correspondingAuthor":false,"prefix":"","firstName":"Hening","middleName":"","lastName":"Li","suffix":""},{"id":436214970,"identity":"ce48b993-baf3-4a63-94c5-cdf87e1dfeb6","order_by":1,"name":"Shanhang Li","email":"","orcid":"","institution":"Guangxi Medical University Cancer Hospital","correspondingAuthor":false,"prefix":"","firstName":"Shanhang","middleName":"","lastName":"Li","suffix":""},{"id":436214971,"identity":"70f4b87b-de69-4fa1-b8e2-3d2168599139","order_by":2,"name":"Liang Xiong","email":"","orcid":"","institution":"the First Affiliated Hospital of Guangxi Medical University","correspondingAuthor":false,"prefix":"","firstName":"Liang","middleName":"","lastName":"Xiong","suffix":""},{"id":436214972,"identity":"f1a8e73b-40aa-4e74-bc1d-e7581204f584","order_by":3,"name":"Mingxiu Yang","email":"","orcid":"","institution":"the First Affiliated Hospital of Guangxi Medical University","correspondingAuthor":false,"prefix":"","firstName":"Mingxiu","middleName":"","lastName":"Yang","suffix":""},{"id":436214973,"identity":"ec113270-4c28-4ca2-8fec-eccf277fe9a7","order_by":4,"name":"Wei Dai","email":"","orcid":"","institution":"the First Affiliated Hospital of Guangxi Medical University","correspondingAuthor":false,"prefix":"","firstName":"Wei","middleName":"","lastName":"Dai","suffix":""},{"id":436214974,"identity":"bc501b81-00ed-4b0c-a279-3c45b0ce77ad","order_by":5,"name":"Xiaoting Luo","email":"","orcid":"","institution":"the First Affiliated Hospital of Guangxi Medical University","correspondingAuthor":false,"prefix":"","firstName":"Xiaoting","middleName":"","lastName":"Luo","suffix":""},{"id":436214975,"identity":"a844a303-39b0-41dd-a444-f540a6fbbcaf","order_by":6,"name":"Shangyu Liu","email":"","orcid":"","institution":"the First Affiliated Hospital of Guangxi Medical University","correspondingAuthor":false,"prefix":"","firstName":"Shangyu","middleName":"","lastName":"Liu","suffix":""},{"id":436214976,"identity":"b27ca696-af64-462f-96cb-663d148d8860","order_by":7,"name":"Danting Xiao","email":"","orcid":"","institution":"the First Affiliated Hospital of Guangxi Medical University","correspondingAuthor":false,"prefix":"","firstName":"Danting","middleName":"","lastName":"Xiao","suffix":""},{"id":436214977,"identity":"2ff3f822-adf7-41ad-80a9-d776248e8845","order_by":8,"name":"Binyuan Ning","email":"","orcid":"","institution":"the First Affiliated Hospital of Guangxi Medical University","correspondingAuthor":false,"prefix":"","firstName":"Binyuan","middleName":"","lastName":"Ning","suffix":""},{"id":436214978,"identity":"50eb696f-96db-4152-aaa9-3736cac408e2","order_by":9,"name":"Yun Liu","email":"","orcid":"","institution":"the First Affiliated Hospital of Guangxi Medical University","correspondingAuthor":false,"prefix":"","firstName":"Yun","middleName":"","lastName":"Liu","suffix":""},{"id":436214979,"identity":"999fc55c-c953-4c69-811b-97d1351ce3d6","order_by":10,"name":"Haijun Tang","email":"","orcid":"","institution":"the First Affiliated Hospital of Guangxi Medical University","correspondingAuthor":false,"prefix":"","firstName":"Haijun","middleName":"","lastName":"Tang","suffix":""},{"id":436214980,"identity":"bb81aca0-8e62-4ae8-9b4f-33544a0052ec","order_by":11,"name":"Juliang He","email":"","orcid":"","institution":"Guangxi Medical University Cancer Hospital","correspondingAuthor":false,"prefix":"","firstName":"Juliang","middleName":"","lastName":"He","suffix":""},{"id":436214985,"identity":"77540f88-b77f-43e6-bf90-6329b00c3414","order_by":12,"name":"Jie Ma","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAwklEQVRIiWNgGAWjYHCChAMMBjZy/MzMhx+QoKUizViynS3NgASLzhxONDjPoyBBlGLd9gMPD3xsS0swPszDYMBQYxNNUIvZmYSEgzPbbPLMDvMeeMBwLC23gaCWAwkJh3nb0orNDvMlGDA2HCZCy/kHCYf/th1O3NzMYyBBnJYbQFtA3t/ATLyWBwkHe4CBLHEYGMgJRPnlfE7yhx+gqOw/fPjBhxobwloYGHgSEOwEXIpQAfsB4tSNglEwCkbByAUASEFHg2FK18QAAAAASUVORK5CYII=","orcid":"","institution":"the First Affiliated Hospital of Guangxi Medical University","correspondingAuthor":true,"prefix":"","firstName":"Jie","middleName":"","lastName":"Ma","suffix":""}],"badges":[],"createdAt":"2025-03-27 06:23:22","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6317412/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6317412/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":79819447,"identity":"e10e2fda-4a8d-4de9-9624-b5602a85a20a","added_by":"auto","created_at":"2025-04-03 08:27:44","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":14685110,"visible":true,"origin":"","legend":"\u003cp\u003eSingle-cell transcriptomic analysis of PD-1 blockade therapy. (A) Schematic overview of the study design (created using BioRender.com). Detailed information about cohorts and samples is provided in Supplementary Tables 1 and 2. (B) Uniform Manifold Approximation and Projection (UMAP) plot displaying major cell types. Each dot represents a single cell, and colors indicate different cell populations. (C) Marker gene expression for cell clusters. Bubble size is proportional to the percentage of cells expressing the gene, while color intensity reflects the average scaled gene expression. (D) Proportions of different cell types in response (R) and non-response (NR) groups. (E) Distribution of PDCD1 (PD-1) and CD274 (PD-L1) expression levels and proportions across different cell types in the R and NR groups.\u003c/p\u003e","description":"","filename":"Figure1.png","url":"https://assets-eu.researchsquare.com/files/rs-6317412/v1/f3823a87161b53e0849ff4c1.png"},{"id":79819451,"identity":"6f653df6-0f5e-4d5c-ad28-cc64a1dce149","added_by":"auto","created_at":"2025-04-03 08:27:44","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":26810667,"visible":true,"origin":"","legend":"\u003cp\u003eTranscriptomic diversity of T cells. (A) UMAP plots showing the clustering of nine T-cell subsets (left) and their distribution based on treatment response (right). (B) Expression of marker genes in the defined T-cell clusters. (C) Heatmap displaying the expression of 19 selected gene signatures across T-cell clusters. The heatmap is generated based on scaled gene signature scores. (D) Bubble plot illustrating the expression of key marker genes among three T-cell clusters. \u0026nbsp;(E) Response preference of each T-cell cluster, estimated by Ro/e. (F) UMAP plots showing T-cell density in R and NR groups.\u003c/p\u003e","description":"","filename":"Figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-6317412/v1/bc2218cffe961fb5f5b26063.png"},{"id":79819450,"identity":"43225f98-161f-4122-aa11-a7cdb662aa7a","added_by":"auto","created_at":"2025-04-03 08:27:44","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":26853636,"visible":true,"origin":"","legend":"\u003cp\u003eDifferentiation and functional states of T cells. (A) Differentiation trajectory of T cells, with cells color-coded according to their corresponding pseudotime. (B) Expression dynamics of four representative gene signatures along pseudotime. (C-D) Two-dimensional plots illustrating the dynamic expression of selected genes along pseudotime (C), with coloring based on cell clusters (D). (E) Distribution of CD8+ effector T cells along pseudotime in R and NR groups. (F) Enrichment of cytotoxic T-cell signatures and T-cell-mediated tumor immune response features across three T-cell clusters. Statistical significance was determined by the Wilcoxon test (**p \u0026lt; 0.01, **p \u0026lt; 0.0001).\u003c/p\u003e","description":"","filename":"Figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-6317412/v1/b3261aa233554711008e7f30.png"},{"id":79819459,"identity":"58a95396-45d8-4319-82e1-1486cfc2f9a8","added_by":"auto","created_at":"2025-04-03 08:27:44","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":24463080,"visible":true,"origin":"","legend":"\u003cp\u003eIdentification of CD8_Teff cells. (A) Heatmap showing the proportions of T-cell subpopulations across different cancer types. (B) Proportion of CD8_Teff cells across various cancer types. (C) Volcano plot illustrating DEGs between CD8_Teff cells in the R and NR groups. (D) Enriched pathways in CD8_Teff cells were identified by comparing the R group to the NR group. (E) Heatmap of five gene modules. (F) Correlation between gene modules and response groups. (G) Volcano plot displaying DEGs between responders and non-responders in bulk RNA-seq data. (H) Venn diagram showing the overlap between genes in modules (ME1) and upregulated genes in bulk RNA-seq data.\u003c/p\u003e","description":"","filename":"Figure4.png","url":"https://assets-eu.researchsquare.com/files/rs-6317412/v1/5ffc5122ee8e59aebed3434b.png"},{"id":79819453,"identity":"edf5463e-54ed-46b7-a5b5-29445040a4a3","added_by":"auto","created_at":"2025-04-03 08:27:44","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":20599591,"visible":true,"origin":"","legend":"\u003cp\u003eETIRGS accurately predicts immune responses to PD-1 blockade therapy. (A) Top 50 machine learning models ranked by performance. (B-C) Gene contributions to the ETIRGS model based on weights (B) and SHAP values (C). The importance of each gene in the ETIRGS model is determined by its weight, while SHAP values represent each gene’s contribution to the predictive performance of the model. (D) Bubble plot showing the expression levels of ETIRGS genes in CD8_Teff cells from the R and NR groups. (E) Enrichment levels of 22 immune cell types across high- and low-prediction score groups. (F) Violin plots comparing stromal scores, immune scores, and ESTIMATE scores between high- and low-prediction score groups.\u003c/p\u003e","description":"","filename":"Figure5.png","url":"https://assets-eu.researchsquare.com/files/rs-6317412/v1/4237a385381996fc9d71018c.png"},{"id":79819838,"identity":"4c8a2bca-b17c-4f1e-bfdd-ec898ece22a0","added_by":"auto","created_at":"2025-04-03 08:35:44","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":36816,"visible":true,"origin":"","legend":"\u003cp\u003eInteractions between macrophages and CD8_Teff cells. (A-B) The number of significant ligand-receptor pairs between CD8_Teff cells and macrophage subpopulations in different response groups. (C-D) Violin plots illustrating the M1 and M2 phenotypes across various myeloid cell subpopulations. Significant differences between groups were confirmed by Kruskal-Wallis tests. (E-F) Bubble plots displaying ligand-receptor interactions mediated by CD8_Teff cells as ligand cells with other myeloid cell subpopulations.\u003c/p\u003e","description":"","filename":"Figure6.png","url":"https://assets-eu.researchsquare.com/files/rs-6317412/v1/4abce14c1c04e89094906103.png"},{"id":92949485,"identity":"c1ce4fba-abcb-483b-9b36-14be935a5454","added_by":"auto","created_at":"2025-10-07 13:04:21","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":112165917,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6317412/v1/42c64147-7346-4d1a-9266-28c8decfe17d.pdf"},{"id":79819446,"identity":"eb29c009-704a-43b2-89d4-16040c99d542","added_by":"auto","created_at":"2025-04-03 08:27:44","extension":"xlsx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":33972,"visible":true,"origin":"","legend":"","description":"","filename":"TableS1SampleinformationofscRNAseqdata.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-6317412/v1/577b1f260a36b50e3597ef21.xlsx"},{"id":79819448,"identity":"acf6a570-d1ef-41d4-aab4-48a9885fe038","added_by":"auto","created_at":"2025-04-03 08:27:44","extension":"xlsx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":10709,"visible":true,"origin":"","legend":"","description":"","filename":"TableS2SampleinformationofbulkRNAseqdata.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-6317412/v1/aec3c5d5fd4e10bcbb1420dc.xlsx"},{"id":79819458,"identity":"57d1414a-85b7-4edb-8e9b-163a6fedaa5e","added_by":"auto","created_at":"2025-04-03 08:27:44","extension":"pdf","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":5452540,"visible":true,"origin":"","legend":"","description":"","filename":"FigureS1S4.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6317412/v1/257d26419a8aa1aae144e300.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Constructing a Predictive Model for PD-1 Blockade Therapy in Pan-Cancer Based on Machine Learning","fulltext":[{"header":"Introduction","content":"\u003cp\u003eImmune checkpoint blockade (ICB) has brought a historic shift in cancer therapy, transforming therapeutic strategies for various solid tumors.\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u003c/sup\u003e Therapeutic inhibitors targeting cytotoxic T-lymphocyte-associated protein 4 (CTLA-4), programmed cell death protein-1 (PD-1), and its ligand PD-L1 have proven effective in reactivating dysfunctional cytotoxic CD8\u0026thinsp;+\u0026thinsp;T cells, thereby enabling the elimination of cancer cells.\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e,\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e Studies show that elevated PD-L1 expression in tumors improves the clinical outcomes of PD-1 blockade therapy.\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e,\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e However, more than half of patients fail to respond to PD-1 blockade therapy, even when combined with other treatments.\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e,\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e Additionally, the therapeutic efficacy of anti-PD-1 treatment varies widely across different cancer types and among individual patients.\u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e Therefore, developing biomarkers to accurately predict clinical outcomes across cancer patients remains an urgent challenge.\u003c/p\u003e \u003cp\u003eCytotoxic CD8\u0026thinsp;+\u0026thinsp;T cells are key effector cells responsible for the effectiveness of anti-PD-1 therapy, and their infiltration into the tumor microenvironment (TME) closely associated with treatment response. The interaction between PD-1 and PD-L1 reduces the activation of effector T cells, inhibiting anti-tumor immunity. In the immunosuppressive TME, infiltrating cytotoxic T cells typically exhibit exhaustion or functional impairment, making it difficult to control tumor growth.\u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e,\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e Blockade of the PD-1/PD-L1 axis restores the cytotoxic potential of effector T cells and induce tumor regression.\u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e,\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e The success of PD-1 blockade therapy depends on the presence of functional cytotoxic effector T cells in the TME. These cells exert their effects through cytotoxic molecules like granzyme and perforin.\u003csup\u003e\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e Thus, a deeper understanding of T cell functions in the TME is critical for uncovering the mechanisms underlying immune therapy and identifying predictive biomarkers.\u003c/p\u003e \u003cp\u003eIn the context of anti-tumor immunity, biomarker development has become essential for predicting therapeutic responses. While commonly used biomarkers such as PD-L1, tumor mutational burden (TMB), circulating tumor DNA (ctDNA), gut microbiota, and dMMR/MSI-H\u003csup\u003e13\u0026ndash;16\u003c/sup\u003e have shown some predictive value in clinical practice, they still have limitations. Their effectiveness is often constrained by technical and biological factors.\u003csup\u003e\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e,\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u003c/sup\u003e To enhance the universality and accuracy of biomarkers, there is an urgent need to develop more objective and feasible tools to predicting treatment outcomes. Recently, several studies have leveraged genomic and transcriptomic data from tumor patients\u003csup\u003e\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e,\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u003c/sup\u003e to identify multiple biomarkers that can better predict responses to ICB therapy.\u003c/p\u003e \u003cp\u003eRecent advancements in single-cell RNA sequencing (scRNA-seq) have allowed detailed profiling of tumor-infiltrating immune cells, uncovering diverse T cell subpopulations within the TME.\u003csup\u003e\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e,\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u003c/sup\u003e Simultaneously, machine learning (ML), renowned for its exceptional performance in data mining, has been widely employed across various research domains. The integration of scRNA-seq with ML has not only significantly enhanced the efficiency of mining high-dimensional data but also facilitated the precise identification of specific immune cell subpopulations, the screening of potential biomarkers, and the prediction of therapeutic responses. This study integrated scRNA-seq and bulk RNA-seq to analyze the immune landscape linked to treatment response, highlighting a significant correlation between CD8 effector T cells and immune therapy response. Additionally, LASSO regression analysis was used to select genes associated with immune response, and 152 predictive models were constructed using 15 ML algorithms to predict immune therapy response. A novel effector T cell immunotherapy response genes (ETIRGS) was developed through multi-cohort validation, offering a new framework for predicting PD-1 blockade efficacy. This study underscores the importance of CD8 effector T cells in anti-tumor immunity across various cancers and offers a promising predictive biomarker for optimizing immune therapeutic strategies.\u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003ePan-cancer Single-cell Expression Atlas of Anti-PD-1 Therapy\u003c/h2\u003e \u003cp\u003eTo assess the tumor immune characteristics across various cancer types, we performed scRNA-seq analysis on 293 samples from 159 patients with seven different cancer types, aiming to construct a single-cell atlas for in-depth analysis (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eA; Figure \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003eA-B). Each cancer type was associated with corresponding treatment response data (Supplementary Table\u0026nbsp;1). After rigorous quality control filtering, a single-cell expression atlas consisting of 751,178 cells was generated. Using well-defined marker genes, the cells were categorized into eight major cell types: five immune cell populations (T cells, NK cells, B cells, myeloid cells, and mast cells) and three stromal cell populations (endothelial cells, epithelial cells, and fibroblasts) (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eB-C; Figure \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003eC-D). Furthermore, the relative percentages of each cell type in the TME across different groups were displayed as bar plots. The results showed a significant increase in the proportions of T cells, B cells, and NK cells in the response (R) group (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eD). PD-1 (PDCD1) expression was primarily confined to T cells and NK cells, with significantly higher expression in the R group compared to the non-response (NR) group. Conversely, PD-L1 (CD274) was highly expressed in myeloid cells and mast cells (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eE).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThese findings highlight the complex and heterogeneous cellular composition of the TME in pan-cancer immunotherapy and its relationship with treatment response.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eT-cell Subpopulation Characteristics and Significant Differences in Response Status\u003c/h3\u003e\n\u003cp\u003eIt is well-established that T cell infiltration within tumors plays a central role in PD-1 blockade therapy.\u003csup\u003e\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e,\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e Using our high-resolution T-cell atlas across pan-cancer, we identified nine major T-cell subpopulations, including naive T cells (Tn), stress response CD8\u0026thinsp;+\u0026thinsp;T cells (CD8_Tstr), interferon (IFN) response CD8\u0026thinsp;+\u0026thinsp;T cells (CD8_Tisg), type 17 helper T(Th17), exhausted CD4\u0026thinsp;+\u0026thinsp;T cells (CD4_Tex), regulatory T cells (CD4_Treg), transitional effector CD8\u0026thinsp;+\u0026thinsp;T cells (CD8_tTeff), effector CD8\u0026thinsp;+\u0026thinsp;T cells (CD8_Teff), and natural killer T cells (NKT) (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA, Figure \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003eA). Among these, Tn exhibited typical naive signatures with high expression of naive genes (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eB-C). CD8_Tstr and CD8_Tisg exhibited molecular phenotypes shaped by environmental factors. Specifically, CD8_Tstr was characterized by high expression of stress-related heat shock genes (such as HSPA1B and HSPA1A),\u003csup\u003e\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e,\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u003c/sup\u003e along with unique stress response gene signatures, while CD8_Tisg exhibited high levels of IFN-stimulated genes and IFN-response signatures (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eB-C). Previous studies have shown that Tstr cells can be detected in situ within the TME of various cancer types, and stress-related genes are significantly upregulated in non-responsive tumors following immune checkpoint blockade therapy.\u003csup\u003e\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e CD4_Tex exhibited high expression of exhaustion markers (such as PDCD1, TIGIT and CTLA4) and transcription factors (TFs) (such as TOX and TOX2), along with upregulated exhaustion-related gene signatures and MARK signaling pathways (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eB-C). CD4_Treg exhibited classic Treg gene features, including IL2RA, FOXP3, TNFRSF4, and CTLA4. CD8_tTeff was positioned between activation and cytotoxicity, with elevated expression of glycolysis-related genes (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eB-D). CD8_Teff cells showed elevated expression of granzyme genes (GZMA, GZMB, GZMK, and NKG7), IFN-γ (IFNG), Fas ligand (FASLG), and chemokines (CCL4 and CCL5), with a strong enrichment of CD8 cytotoxic gene signatures (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eB-D), indicating their effector functions. In contrast to CD8_Teff, NKT cells displayed stronger expression of NK-associated genes (KLRD1, FCGR3A, KLRF1), along with high expression of fatty acid metabolism and pro-apoptotic genes (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eB-D). Moreover, we observed increased infiltration of CD8_Teff and CD4_Tex in R groups, while infiltration of CD8_Tisg was decreased (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eE-F).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eTo better understand the differentiation trajectories of T cells under anti-PD-1 therapy, we performed Monocle3 analysis, which identified three main branches. All branches originated from Tn cells. Path 1 terminated in CD4_Tex and CD4_Treg, while paths 2 and 3 passed through CD8_Tstr, with path 2 ending in CD8_Teff and NKT, and path 3 concluding with CD8_Tisg (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA). These pathways represent distinct cellular fates, supported by the expression dynamics of relevant gene sets inferred along the pseudotime axis (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eB). Effector activation and cytotoxic signatures increased along the pseudotime trajectory of CD8_Teff cells, while exhaustion markers declined. Differential gene expression analysis showed significant upregulation of cytotoxic-related genes (GZMA, GZMB, GZMH, GNLY) and chemokines (CCL5, CCL4) along the pseudotime progression in the developmental lineage of CD8_Teff (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eC-D). Furthermore, the developmental state analysis of T cells showed that CD8_Teff from responders were positioned later in the lineage, while CD8_Teff from non-responders were at an earlier stage of development. No significant differences were observed in other T cell subpopulations (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eE, Figure \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003eB). This suggests that CD8_Teff from responders is more fully differentiated than those from non-responders. CD8_Teff cells in the R group displayed higher cytotoxic characteristics and stronger T cell-mediated anti-tumor activity (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eF).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThese results underscore the pivotal role of T-cell subpopulations, particularly CD8_Teff, in the efficacy of anti-PD-1 therapy.\u003c/p\u003e\n\u003ch3\u003eActivated and exhausted-like CD8_Teff Are Associated with Response\u003c/h3\u003e\n\u003cp\u003eThis study the composition of T-cell subpopulations across different cancer types was analyzed in detail (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eA). Notably, CD8_Teff comprised a significant proportion of T cells, and its proportion varied notably across different tumor types. For example, the highest proportion of CD8_Teff was observed in renal clear cell carcinoma (RCC), while the lowest was observed in basal cell carcinoma (BCC) (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eB). This suggests that CD8_Teff may exhibit varying degrees of immune activation enrichment in different cancer types. We further investigated the distribution of CD8_Teff across various tumor tissues. In most cancer types, a higher proportion of CD8_Teff was found in the R group (Figure \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003eC), which further supports the pivotal role of CD8_Teff in regulating anti-tumor immune responses. Interestingly, within the tumor tissues, CD8_Teff in the R group exhibited higher expression of chemokines (such as CXCL13, CXCR6, and CCL4L2), cytotoxic genes (GZMB, GZMA), and exhaustion markers (PDCD1, CTLA4) compared to the NR group (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eC). The elevated expression of these genes is strongly linked to immune regulatory pathways, including T-cell activation, antigen processing, and presentation. In contrast, the NR group displayed upregulation of pathways associated with cell division (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eD).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003ePD-1 is a key marker of tumor-reactive T cells in human cancers and is considered a potential biomarker for tumor recognition rather than effector dysfunction.\u003csup\u003e\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e,\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e\u003c/sup\u003e High PD-1 expression in CD8\u0026thinsp;+\u0026thinsp;T cells, which exhibit both activation and exhaustion-like phenotypes, is critical for immune checkpoint blockade responses due to their potent anti-tumor activity. \u003csup\u003e\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e,\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e\u003c/sup\u003e Additionally, PD-1 blockade restores cytotoxic function in PD-1\u0026thinsp;+\u0026thinsp;CD8\u0026thinsp;+\u0026thinsp;T cells and enhances immunosuppressive activity in PD-1\u0026thinsp;+\u0026thinsp;Treg cells.\u003csup\u003e\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e\u003c/sup\u003e In this study, PD-1 expression was elevated in CD8_Teff cells in the R group, while it was lower in CD4_Treg cells (Figure \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003eD). Furthermore, the ratio of PD-1\u0026thinsp;+\u0026thinsp;CD8_Teff to PD-1\u0026thinsp;+\u0026thinsp;CD4_Treg was significantly higher in the R group than in the NR group (Figure \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003eE). This finding indicates that the PD-1\u0026thinsp;+\u0026thinsp;CD8\u0026thinsp;+\u0026thinsp;T-cell to PD-1\u0026thinsp;+\u0026thinsp;CD4_Treg ratio in the TME is closely associated with the efficacy of immune checkpoint therapy. The ratio of PD-1\u0026thinsp;+\u0026thinsp;CD8\u0026thinsp;+\u0026thinsp;T cells to PD-1\u0026thinsp;+\u0026thinsp;Treg cells is considered a potential predictive marker for PD-1 blockade efficacy, potentially surpassing other factors like PD-L1 expression and tumor mutation burden.\u003csup\u003e\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e\u003c/sup\u003e In conclusion, these results highlight the potential of CD8_Teff as a critical biomarker for the effectiveness of immune therapies.\u003c/p\u003e \u003cp\u003eNext, Hotspot\u003csup\u003e\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e\u003c/sup\u003e was used to to analyze DEGs (log2FC\u0026thinsp;\u0026gt;\u0026thinsp;0 and p_val_adj\u0026thinsp;\u0026lt;\u0026thinsp;0.05) in the CD8_Teff dataset, aiming to identify gene sets closely linked to immune therapy. Five gene modules were identified (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eE). Figure \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003eF shows the top ten genes for each module. All five modules exhibited significant differences between the R and NR groups. Module 1 demonstrated the strongest correlation in the R group, making it the most representative module (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eF). Consequently, module 1 were selected for further analysis, totaling 88 genes.\u003c/p\u003e\n\u003ch3\u003eDevelopment and Validation of the ETIRGS\u003c/h3\u003e\n\u003cp\u003eTo further investigate the immune microenvironment in PD-1 blockade therapy, the RNA-seq expression profiles of 172 pan-cancer bulk samples were analyzed. A volcano plot was generated to display DEGs between the R and NR groups, identifying 635 up-regulated and 223 down-regulated genes (log2FC\u0026thinsp;\u0026gt;\u0026thinsp;1 or \u0026lt; -1, p_val_adj\u0026thinsp;\u0026lt;\u0026thinsp;0.05) (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eG). Given the pivotal role of CD8_Teff in PD-1 blockade therapy, we initially intersected the 88 genes from module 1 with the 635 up-regulated genes, resulting in 16 candidate genes (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eH). These genes were further optimized using LASSO regression analysis, identifying 7 ETIRGS: CXCL13, GZMB, KLRD1, PDCD1, LSP1, BATF and ABI3. Based on these 7 genes, we constructed 152 ML prediction models using 15 algorithms. The models were compared by calculating their F-scores, identifying the best model (NN-MLP\u0026thinsp;+\u0026thinsp;Lasso-CV; parameter settings: cutoff: 0.25, lr: 0.001, bs: 115, ep: 50, dropout: 0.5) (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eA; Figure \u003cspan refid=\"MOESM3\" class=\"InternalRef\"\u003eS3\u003c/span\u003eA-B). To further assess each gene's contribution to the model, we compared their weights and evaluated their importance using the SHAP framework. The results indicated that CXCL13 and GZMB were identified as top contributors with higher model weights. (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eB-C). Correlation analysis of the 7 ETIRGS revealed positive associations, with CXCL13 showing the strongest correlation with GZMB and PDCD1. (Figure \u003cspan refid=\"MOESM3\" class=\"InternalRef\"\u003eS3\u003c/span\u003eC). Notably, the expression of all 7 ETIRGS was significantly higher in the R group than in the NR group (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eD), indicating that these genes could serve as independent predictive biomarkers for PD-1 blockade efficacy.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eImmune cell infiltration analysis using the CIBERSORT algorithm revealed significantly higher levels of CD8 T cells and M1 macrophages in the high-predicted scoring group compared to the low-predicted group (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eE). Additionally, stromal, immune, and ESTIMATE scores, evaluated through the ESTIMATE function, were also significantly elevated in the high-predicted group (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eF). This further supports the potential of ETIRGS as reliable predictive biomarkers for the efficacy of PD-1 blockade therapy.\u003c/p\u003e\n\u003ch3\u003eImpact of Macrophage Phenotypes and CD8_Teff Interaction on Treatment Efficacy\u003c/h3\u003e\n\u003cp\u003eMyeloid cells are pivotal in tumor immunity.\u003csup\u003e\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e,\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e\u003c/sup\u003e In this study, myeloid cells in the pan-cancer dataset were subdivided into 8 distinct clusters (Figure S4A-B). To explore whether interactions between CD8_Teff and myeloid cells contribute to immune therapy response, using the CellChat method, it was found that ligand-receptor interactions between CD8_Teff and CXCL10\u0026thinsp;+\u0026thinsp;macrophages were significantly enriched in the R group, while the NR group showed stronger interactions between CD8_Teff and SPP1\u0026thinsp;+\u0026thinsp;macrophages (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eA-B). Macrophages can polarize into pro-inflammatory M1 or anti-inflammatory M2 phenotypes, thereby modulating T cell responses in anti-tumor immunity.\u003csup\u003e\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e,\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e\u003c/sup\u003e Further analysis revealed that CXCL10\u0026thinsp;+\u0026thinsp;macrophages displayed M1 characteristics, while SPP1\u0026thinsp;+\u0026thinsp;macrophages were predominantly polarized to the M2 phenotype (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eC-D, Figure S4C-D). CXCL10 is a well-known chemokine that mediates T cell recruitment.\u003csup\u003e\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e,\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e\u003c/sup\u003e The high expression of CXCL10 in CXCL10\u0026thinsp;+\u0026thinsp;macrophages supports their pro-inflammatory properties, suggesting that the pro-inflammatory macrophage phenotype is likely shaped by active anti-tumor immune responses within the TME. Previous studies have shown that M2-like SPP1\u0026thinsp;+\u0026thinsp;macrophages are closely linked to angiogenesis and poor prognosis in patients.\u003csup\u003e\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eSubsequently, we specifically focused on ligand-receptor interactions. In the R group, the number of ligand-receptor interactions between CD8_Teff and CXCL10\u0026thinsp;+\u0026thinsp;macrophages was the highest among the myeloid cell subpopulations. Importantly, CD8_Teff cells produced multiple ligands (CCL5 and CCL4), which have significant regulatory potential. These ligands bind to the receptors (CCR5 and CCR1) on CXCL10\u0026thinsp;+\u0026thinsp;macrophages, promoting the migration of pro-inflammatory macrophages to the tumor site, thereby cooperating to drive the immune response (Figs.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eE-F). Overall, in responsive tumors, CD8_Teff cells likely shape the phenotype of pro-inflammatory macrophages. The pro-inflammatory signaling molecules expressed by CD8_Teff and CXCL10\u0026thinsp;+\u0026thinsp;macrophages may further enhance anti-tumor immunity. In contrast, the M2-type SPP1\u0026thinsp;+\u0026thinsp;macrophages contribute to an immune-suppressive TME.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003ePD-1 blockade therapy has shown broad clinical potential across various tumor types.\u003csup\u003e\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e\u003c/sup\u003e The critical role of tumor-infiltrating T cells in immune responses is well-established.\u003csup\u003e\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e\u003c/sup\u003e However, a universal biomarker to assess immune therapy responsiveness across cancer types is still lacking. Therefore, there is an urgent need to develop a universal metric to accurately evaluate immune therapy efficacy across cancer types.\u003c/p\u003e \u003cp\u003eTo address this gap, we integrated gene expression profiles from diverse tumor samples to explore the heterogeneity of tumor-infiltrating T cells across cancers and identify key immune features linked to immune therapy response. Our findings identify that CD8_Teff can serve as a potential predictive biomarker for PD-1 blockade therapy. The differentiation status, infiltration levels, and high expression of effector activation and cytotoxicity genes in CD8_Teff correlate closely with treatment response. Furthermore, we found that CD8_Teff from responders strongly interact with pro-inflammatory macrophages (CXCL10\u0026thinsp;+\u0026thinsp;macrophages) in the TME, primarily through ligand-receptor pairs such as CCL5-CCR5, CCL5-CCR1, and CCL4-CCR5. These cells are co-enriched in the TME of responders, suggesting that the significant response to PD-1 blockade therapy in tumor patients may be driven by the interaction between CD8 effector T cells and pro-inflammatory macrophages, thereby enhancing anti-tumor immunity. Next, using Lasso regression, we applied 15 ML methods to fit 152 models, ultimately establishing the ETIRGS model, which shows significant potential in predicting immune therapy efficacy. The optimal model was derived using the NN-MLP\u0026thinsp;+\u0026thinsp;Lasso-CV algorithm. Compared to traditional single biomarkers, the ETIRGS model offers superior and more stable performance. As an independent indicator, it effectively complements and overcomes the limitations of traditional biomarkers. These results provide valuable insights for future studies on immune therapy responsiveness.\u003c/p\u003e \u003cp\u003eThe ETIRGS comprises of seven key genes: CXCL13, GZMB, KLRD1, PDCD1, LSP1, BATF and ABI3. Several of these genes have been identified as potential predictive biomarkers for ICB therapy. Among these, CXCL13 holds substantial academic value for exploration. Originally termed B-cell attracting chemokine 1 (BCA-1) due to its strong chemotactic effect on B cells,\u003csup\u003e\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e,\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e\u003c/sup\u003e recent studies revealed that CXCL13 is secreted by diverse T cell populations, such as T follicular helper cells (TFH), tumor-infiltrating CD8\u0026thinsp;+\u0026thinsp;T cells, and exhausted CD8\u0026thinsp;+\u0026thinsp;T cells.\u003csup\u003e\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e\u003c/sup\u003e In multiple cancer types, CXCL13 expression closely correlates with ICB response.\u003csup\u003e\u003cspan additionalcitationids=\"CR45\" citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e\u003c/sup\u003e For instance, in non-small cell lung cancer (NSCLC), CXCL13 expression was observed in PD-1\u003csup\u003ehigh\u003c/sup\u003e cytotoxic T cells, highlighting its potential as a biomarker for PD-1 blockade therapy response.\u003csup\u003e\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u003c/sup\u003e In triple-negative breast cancer (TNBC) patients, CXCL13\u0026thinsp;+\u0026thinsp;PD-1\u0026thinsp;+\u0026thinsp;T cells clonally expand significantly after PD-1 blockade therapy, exhibiting cytotoxic activity and exhaustion markers.\u003csup\u003e\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e\u003c/sup\u003e Similarly, CXCL13\u0026thinsp;+\u0026thinsp;T cells were identified in liver metastasis of gallbladder cancer, correlating with higher immune scores.\u003csup\u003e\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e\u003c/sup\u003e Additionally, in EB virus-related gallbladder cancer, PD-1 blockade therapy activated CXCL13-expressing effector cell subsets.\u003csup\u003e\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e\u003c/sup\u003e As a further exploration, our study integrated analyses of various cancer samples and confirmed that CD8_Teff cells from responders exhibit both activation and exhaustion characteristics, capable of activating CXCL10\u0026thinsp;+\u0026thinsp;macrophages into a pro-inflammatory phenotype. Furthermore, CXCL13 expression was positively correlated with inhibitory genes (e.g., PDCD1) and cytotoxic molecules (e.g., GZMB). We propose that CXCL13 may serve as a unifying marker for CD8 effector T cells in the TME, with broad potential to predict responses to PD-1 blockade therapy. The potential role of CXCL13 as a positive predictive biomarker for immune therapy warrants further investigation. Moreover, GZMB, a serine protease, is a key factor mediating apoptosis in target cells by NK cells and cytotoxic CD8 T cells.\u003csup\u003e\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e,\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e\u003c/sup\u003e GZMB\u0026thinsp;+\u0026thinsp;CD8 T cells are significant predictive markers for ICB therapy response and can effectively predict treatment sensitivity.\u003csup\u003e\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e\u003c/sup\u003e Compared to previous studies, the combined prediction model of CXCL13 and GZMB for PD-1 blockade therapy efficacy outperforms single biomarker models, potentially helping to identify patients who are most likely to benefit from immune therapy. Our predictive model strongly supports patient stratification based on immune treatment response, facilitating the development of personalized treatment strategies.\u003c/p\u003e \u003cp\u003eNevertheless, our study has certain limitations. Firstly, this study is retrospective, and all data and clinical information are derived from public databases. Larger sample studies are needed for future validation. Secondly, the role of ETIRGS in excluded solid tumors remains unclear, necessitating further research to validate its universality and clinical applicability across diverse cancer types.\u003c/p\u003e \u003cp\u003eIn conclusion, our study identifies key immune cell subsets associated with the response to PD-1 blockade therapy across various cancers and provides new universal biomarkers for predicting therapeutic efficacy. Further studies are required to validate our findings.\u003c/p\u003e"},{"header":"Material and methods","content":"\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003eData Collection\u003c/h2\u003e \u003cp\u003eThis study included nine scRNA-seq datasets, comprising single-cell transcriptomic data from 258 samples of 159 individuals. These datasets represent seven different cancer types, with clinical information on treatment response status for each cancer type (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eA, Supplementary Table\u0026nbsp;1). Additionally, bulk RNA-seq datasets from eight cancer types treated with anti-PD-1 therapy, comprising 172 samples, were obtained. Only tumor samples obtained post-immunotherapy were retained for analysis. Supplementary Table\u0026nbsp;2 provides the dataset accession numbers and references.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eSingle-Cell Data Processing\u003c/h2\u003e \u003cp\u003eScRNA-seq data integration and analysis were conducted in R (v.4.0.3) using the Seurat (v.4.3) package. The following procedure was adopted: First, the expression matrix of each sample was imported into R with the Read10X() function, and a Seurat object was created, integrating the corresponding clinical information. The Seurat object was then converted into a Scanpy object (Python v.3.9.0) using the SeuratDisk package for further analysis. The Scrublet package (v.0.2.3) was used to predict and filter potential doublets. Quality control was based on the following criteria: (1) total UMI count per cell\u0026thinsp;\u0026lt;\u0026thinsp;25,000; (2) the number of detected genes between 800 and 5,000; (3) excluding cells with 98% of genes expressed; and (4) the percentage of mitochondrial genes\u0026thinsp;\u0026lt;\u0026thinsp;10%. Next, the raw counts were normalized and log-transformed using the Scanpy package (v.1.9.8). After filtering for highly variable genes (HVGs), the first 10 principal components (PCs) were selected for subsequent analysis. The BBKNN algorithm was used to construct a batch-balanced k-nearest neighbor (kNN) graph.\u003csup\u003e\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e\u003c/sup\u003e After batch effect correction, cells were clustered using the Leiden algorithm. Differentially expressed genes (DEGs) were identified using the sc.tl.rank_genes_group() function in Scanpy with the 'use_raw\u0026thinsp;=\u0026thinsp;True' parameter.\u003csup\u003e\u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e54\u003c/span\u003e\u003c/sup\u003e Based on the DEGs of each cluster and known marker genes, major cell types were identified, including T cells (CD3D, CD3E), NK cells (NKG7, KLRD1, NCAM), B cells (CD79A, MZB1), myeloid cells (LYZ, CD68), mast cells (TPSB1, KIT), endothelial cells (VWF, EMCN), epithelial cells (KRT5, KRT14), and fibroblasts (COL6A2, COL1A2). Clusters expressing multiple standard marker genes were considered as doublet cells. Doublets and undefined cells were excluded, resulting in a total of 630,638 cells for further analysis. The Scanpy object was then converted back into a Seurat object using the sceasy package (v.0.0.7). Subclustering of the main cell types was performed based on the original expression matrix, using the same preprocessing steps, which resulted in the identification of subcluster structures.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003eIdentification of T-cell Types and Functional States\u003c/h2\u003e \u003cp\u003eTo determine T-cell types and their functional states, we followed the methodology proposed by Chu et al..\u003csup\u003e\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e First, DEGs for major cell types were identified using the FindAllMarkers() function from the Seurat package. Subsequently, a bubble plot was generated to visualize the DEGs and T-cell marker genes (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eB). Finally, the AddModuleScore() function in Seurat was employed to calculate gene feature scores for selected gene sets linked to functional states within each T-cell cluster.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003eTissue Distribution of Clusters\u003c/h2\u003e \u003cp\u003eTo quantify the tissue preference of each cluster across different immune response groups, the ratio of observed to expected cell counts (Ro/e) was calculated for each cluster.\u003csup\u003e\u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e55\u003c/span\u003e,\u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e56\u003c/span\u003e\u003c/sup\u003e Specifically, a chi-square test was performed to obtain the expected cell counts for each cell cluster and tissue combination. If Ro/e\u0026thinsp;\u0026gt;\u0026thinsp;1, it indicated a higher presence of the cluster in the specific immune response group.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003eSingle-Cell Differentiation Trajectory Inference\u003c/h2\u003e \u003cp\u003eMonocle 3 (v.1.3.7)\u003csup\u003e\u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e57\u003c/span\u003e\u003c/sup\u003e was used to infer the differentiation trajectories of T cells. The expression matrix of the T cell Seurat object was input into Monocle 3, and a cds object was created using the new_cell_data_set() function. Subsequently, dimensionality reduction, cell clustering, and differentiation trajectory inference were performed. Differential gene analysis along the trajectory was conducted using the graph_test() function.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003eGene Feature Calculation\u003c/h2\u003e \u003cp\u003eTwenty-three typical T cell cytotoxicity features were collected based on previous studies on T cells.\u003csup\u003e\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e Genes related to T cell-mediated immune responses against tumor cells were obtained from Li et al.'s study.\u003csup\u003e\u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e58\u003c/span\u003e\u003c/sup\u003e The gene sets for M1 and M2 macrophage signatures were derived from Bi et al.'s study.\u003csup\u003e\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e\u003c/sup\u003e The AddModuleScore() function in the Seurat package was used to score cell subpopulations based on gene sets related to T cell cytotoxicity, T cell-mediated immune responses to tumor cells, and macrophage phenotype. Statistical significance of the scores was assessed using the Wilcoxon test.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003eCell-Cell Interaction Analysis\u003c/h2\u003e \u003cp\u003eCell-cell interactions between T cell subpopulations were systematically inferred and visualized using CellChat (v.1.6.1).\u003csup\u003e\u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e60\u003c/span\u003e\u003c/sup\u003e To explore the impact of PD-1 blockade on intercellular communication network between cell subpopulations, CellChat objects generated using the gene expression matrices from the R and NR groups. The mergeCellChat() function was employed to combine the CellChat objects from both groups, enabling the analysis of differences in the frequency and intensity of inferred intercellular interactions. Subsequently, the netVisual_bubble() function was applied to visualize significant ligand-receptor pair changes in intercellular interactions.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003eHotspot Gene Module Analysis\u003c/h2\u003e \u003cp\u003eHotspot (v.1.1.1) is a tool used to identify co-expressed gene modules in scRNA-seq datasets.\u003csup\u003e\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e\u003c/sup\u003e To identify gene modules associated with immune therapy responses, the DEGs of CD8_Teff cells from the R and NR groups were first extracted. The create_knn_graph() function from the Hotspot package was then used to construct a k-nearest neighbor (KNN) graph, selecting 30 neighboring genes to generate a gene adjacency matrix. Next, compute_autocorrelations() was employed to calculate the autocorrelation of the genes. Genes with an FDR\u0026thinsp;\u0026lt;\u0026thinsp;0.001 were selected for subsequent local correlation calculations. Finally, gene module clustering was performed using the create_modules() function to identify gene modules related to immune therapy response.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003eBulk RNA-seq Analysis\u003c/h2\u003e \u003cp\u003eGene expression data and clinical information for the patients were retrieved from the GEO database (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/geo/\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/geo/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). The raw sequencing data for the samples were aligned and annotated using the GRCh38.p13 reference genome. Raw sequencing data for gastric cancer were downloaded from the European Nucleotide Archive (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ebi.ac.uk/ena/browser/view/PRJNA557841\u003c/span\u003e\u003cspan address=\"https://www.ebi.ac.uk/ena/browser/view/PRJNA557841\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) and subjected to quality control using FastQC, assessing data quality, base distribution, GC content, and adapter contamination. Adapter sequences and low-quality bases were trimmed using the Trim Galore tool. The trimmed reads were then aligned to the GRCh38 reference genome using HISAT2, with the alignment results stored in sorted BAM file format. Gene expression levels were quantified using the featureCounts tool, based on the Gencode v47 genome annotation. The resulting raw count matrix was used for subsequent analyses.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec19\" class=\"Section2\"\u003e \u003ch2\u003eETIRGS Signature Identification\u003c/h2\u003e \u003cp\u003eDifferential expression analysis was performed using the \"limma\" method, with threshold criteria of logFC\u0026thinsp;\u0026gt;\u0026thinsp;1 or \u0026lt; -1 and p_val_adj\u0026thinsp;\u0026lt;\u0026thinsp;0.05. Up-regulated genes were intersected with module genes to identify candidate genes. Subsequently, a rigorous least absolute shrinkage and selection operator (LASSO) regression was conducted using the glmnet (v.4.1.8) package in R to select genes most strongly associated with immune therapy response.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec20\" class=\"Section2\"\u003e \u003ch2\u003eEstablishment of the ETIRGS Model\u003c/h2\u003e \u003cp\u003eIn the model construction process, a total of 15 ML algorithms were employed, including Neural Network (MLP), Logistic Regression, Linear Discriminant Analysis (LDA), Quadratic Discriminant Analysis (QDA), k-Nearest Neighbor (k-NN), Decision Tree, Random Forest, XGBoost Classification, Ridge Regression, Lasso Regression, Elastic Net Regression, Support Vector Machine (SVM), Gradient Boosting Machine (GBM), Stepwise Logistic Regression, and Naive Bayes Algorithm. In total, 152 ML models were generated using these methods. Each algorithm was trained using a 10-fold cross-validation approach on the training dataset to optimize the model. The performance of the models was then evaluated on the test dataset. The accuracy, recall, and F-score for each model were computed and recorded for each fold of the cross-validation. For training the SVM model, a comparison was made among different kernel functions (linear, radial basis, and polynomial), and the best kernel was selected. To enhance model stability and accuracy, an ensemble learning approach was applied, aggregating predictions from different algorithms through a voting mechanism. Specifically, weights were assigned based on each model's performance on the validation set, and a weighted voting strategy was employed to derive the final prediction result. The prediction results of each algorithm were assigned different weights according to their performance in cross-validation, ultimately generating an integrated prediction result to improve the overall model\u0026rsquo;s stability and accuracy.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec21\" class=\"Section2\"\u003e \u003ch2\u003eCIBERSORT\u003c/h2\u003e \u003cp\u003eCIBERSORT was applied to assess differences in immune cell infiltration abundance between high and low ETIRGS prediction score groups. The CIBERSORT () function was employed to quantify the scores for each immune cell subgroup. The relative abundance of all immune cell types in each group was predicted using the leukocyte characteristic matrix (LM22).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec22\" class=\"Section2\"\u003e \u003ch2\u003eStatistical Analysis\u003c/h2\u003e \u003cp\u003eWilcoxon rank-sum test and Kruskal-Wallis test were utilized to compare numerical variables between response groups and across different cell clusters. Statistical significance was defined as p\u0026thinsp;\u0026lt;\u0026thinsp;0.05 for all analyses. Data analysis and visualization were performed using R (v.4.0.3) and Python (v.3.9.0).\u003c/p\u003e \u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgments\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe are grateful to the contributors to the public databases used in this study and all the authors of the study.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eJie Ma, Juliang He, Haijun Tang and Yun Liu contributed to the study design and critical revision of the manuscript. Hening Li, Shanhang Li, Liang Xiong and Mingxiu Yang carried out the study and drafted the manuscript. Wei Dai, Xiaoting Luo, Shangyu Liu, Danting Xiao and Binyuan Ning analyzed the data. All authors read and approved the final manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study was supported by the Natural Science Foundation of Guangxi Zhuang Autonomous Region [Grant No.2025GXNSFAA069103].\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe datasets presented in this study are available in an online repository. The data that support the findings of this study are available from the following databases: GEO database (https://www.ncbi.nlm.nih.gov/geo/), NGDC database (https://ngdc.cncb.ac.cn/omix/), Single Cell Portal (https://singlecell.broadinstitute.org/) and EBI ENA (https://www.ebi.ac.uk/ena/). The accession numbers of the repository can be found in the Supplementary Material.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics approval and informed consent\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study utilized human data obtained from publicly available databases. The study protocol was approved by the Institutional Review Board/Ethics Committee of the First Affiliated Hospital of Guangxi Medical University(Numbers: 2024-E0885).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFootnotes\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eHening Li, Shanhang Li, Liang Xiong and Mingxiu Yang contributed equally to this work.\u003c/p\u003e\n\u003cp\u003eContributor Information\u003c/p\u003e\n\u003cp\u003eJie Ma: [email protected],\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eJuliang He: [email protected],\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eHaijun Tang: [email protected],\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eYun Liu: [email protected].\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eKraehenbuehl, L., Weng, C. H., Eghbali, S., Wolchok, J. D. \u0026amp; Merghoub, T. Enhancing immunotherapy in cancer by targeting emerging immunomodulatory pathways. \u003cem\u003eNat. Rev. Clin. Oncol.\u003c/em\u003e \u003cb\u003e19\u003c/b\u003e, 37\u0026ndash;50. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41571-021-00552-7\u003c/span\u003e\u003cspan address=\"10.1038/s41571-021-00552-7\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSharma, P. \u0026amp; Allison, J. P. Immune checkpoint targeting in cancer therapy: toward combination strategies with curative potential. \u003cem\u003eCell\u003c/em\u003e \u003cb\u003e161\u003c/b\u003e, 205\u0026ndash;214. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.cell.2015.03.030\u003c/span\u003e\u003cspan address=\"10.1016/j.cell.2015.03.030\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZou, W., Wolchok, J. D. \u0026amp; Chen, L. PD-L1 (B7-H1) and PD-1 pathway blockade for cancer therapy: Mechanisms, response biomarkers, and combinations. \u003cem\u003eSci. Transl Med.\u003c/em\u003e \u003cb\u003e8\u003c/b\u003e, 328rv324. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1126/scitranslmed.aad7118\u003c/span\u003e\u003cspan address=\"10.1126/scitranslmed.aad7118\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCarbone, D. P. et al. First-Line Nivolumab in Stage IV or Recurrent Non-Small-Cell Lung Cancer. \u003cem\u003eN Engl. J. Med.\u003c/em\u003e \u003cb\u003e376\u003c/b\u003e, 2415\u0026ndash;2426. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1056/NEJMoa1613493\u003c/span\u003e\u003cspan address=\"10.1056/NEJMoa1613493\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRizvi, N. A. et al. Cancer immunology. Mutational landscape determines sensitivity to PD-1 blockade in non-small cell lung cancer. \u003cem\u003eScience\u003c/em\u003e \u003cb\u003e348\u003c/b\u003e, 124\u0026ndash;128. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1126/science.aaa1348\u003c/span\u003e\u003cspan address=\"10.1126/science.aaa1348\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDarvin, P., Toor, S. M., Sasidharan Nair, V. \u0026amp; Elkord, E. Immune checkpoint inhibitors: recent progress and potential biomarkers. \u003cem\u003eExp. Mol. Med.\u003c/em\u003e \u003cb\u003e50\u003c/b\u003e, 1\u0026ndash;11. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s12276-018-0191-1\u003c/span\u003e\u003cspan address=\"10.1038/s12276-018-0191-1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHe, X. \u0026amp; Xu, C. Immune checkpoint signaling and cancer immunotherapy. \u003cem\u003eCell. Res.\u003c/em\u003e \u003cb\u003e30\u003c/b\u003e, 660\u0026ndash;669. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41422-020-0343-4\u003c/span\u003e\u003cspan address=\"10.1038/s41422-020-0343-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhao, B., Zhao, H. \u0026amp; Zhao, J. Efficacy of PD-1/PD-L1 blockade monotherapy in clinical trials. \u003cem\u003eTher. Adv. Med. Oncol.\u003c/em\u003e \u003cb\u003e12\u003c/b\u003e, 1758835920937612. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1177/1758835920937612\u003c/span\u003e\u003cspan address=\"10.1177/1758835920937612\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMellman, I., Coukos, G. \u0026amp; Dranoff, G. Cancer immunotherapy comes of age. \u003cem\u003eNature\u003c/em\u003e \u003cb\u003e480\u003c/b\u003e, 480\u0026ndash;489. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/nature10673\u003c/span\u003e\u003cspan address=\"10.1038/nature10673\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2011).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZheng, C. et al. Landscape of Infiltrating T Cells in Liver Cancer Revealed by Single-Cell Sequencing. \u003cem\u003eCell\u003c/em\u003e 169, 1342\u0026ndash;1356 e1316, (2017). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.cell.2017.05.035\u003c/span\u003e\u003cspan address=\"10.1016/j.cell.2017.05.035\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDong, H. et al. Tumor-associated B7-H1 promotes T-cell apoptosis: a potential mechanism of immune evasion. \u003cem\u003eNat. Med.\u003c/em\u003e \u003cb\u003e8\u003c/b\u003e, 793\u0026ndash;800. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/nm730\u003c/span\u003e\u003cspan address=\"10.1038/nm730\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2002).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRussell, J. H. \u0026amp; Ley, T. J. Lymphocyte-mediated cytotoxicity. \u003cem\u003eAnnu. Rev. Immunol.\u003c/em\u003e \u003cb\u003e20\u003c/b\u003e, 323\u0026ndash;370. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1146/annurev.immunol.20.100201.131730\u003c/span\u003e\u003cspan address=\"10.1146/annurev.immunol.20.100201.131730\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2002).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang, Y. et al. FDA-Approved and Emerging Next Generation Predictive Biomarkers for Immune Checkpoint Inhibitors in Cancer Patients. \u003cem\u003eFront. Oncol.\u003c/em\u003e \u003cb\u003e11\u003c/b\u003e, 683419. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3389/fonc.2021.683419\u003c/span\u003e\u003cspan address=\"10.3389/fonc.2021.683419\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhou, C. B., Zhou, Y. L. \u0026amp; Fang, J. Y. Gut Microbiota in Cancer Immune Response and Immunotherapy. \u003cem\u003eTrends Cancer\u003c/em\u003e. \u003cb\u003e7\u003c/b\u003e, 647\u0026ndash;660. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.trecan.2021.01.010\u003c/span\u003e\u003cspan address=\"10.1016/j.trecan.2021.01.010\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAhn, S. et al. Changes and prognostic values of tumor-infiltrating lymphocyte subsets after primary systemic therapy in breast cancer. \u003cem\u003ePLoS One\u003c/em\u003e. \u003cb\u003e15\u003c/b\u003e, e0233037. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1371/journal.pone.0233037\u003c/span\u003e\u003cspan address=\"10.1371/journal.pone.0233037\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang, Q. et al. Prognostic and Predictive Impact of Circulating Tumor DNA in Patients with Advanced Cancers Treated with Immune Checkpoint Blockade. \u003cem\u003eCancer Discov\u003c/em\u003e. \u003cb\u003e10\u003c/b\u003e, 1842\u0026ndash;1853. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1158/2159-8290.CD-20-0047\u003c/span\u003e\u003cspan address=\"10.1158/2159-8290.CD-20-0047\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDoroshow, D. B. et al. PD-L1 as a biomarker of response to immune-checkpoint inhibitors. \u003cem\u003eNat. Rev. Clin. Oncol.\u003c/em\u003e \u003cb\u003e18\u003c/b\u003e, 345\u0026ndash;362. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41571-021-00473-5\u003c/span\u003e\u003cspan address=\"10.1038/s41571-021-00473-5\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZouein, J., Kesrouani, C. \u0026amp; Kourie, H. R. PD-L1 expression as a predictive biomarker for immune checkpoint inhibitors: between a dream and a nightmare. \u003cem\u003eImmunotherapy\u003c/em\u003e \u003cb\u003e13\u003c/b\u003e, 1053\u0026ndash;1065. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.2217/imt-2020-0336\u003c/span\u003e\u003cspan address=\"10.2217/imt-2020-0336\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBai, X. et al. Development and validation of a genomic mutation signature to predict response to PD-1 inhibitors in non-squamous NSCLC: a multicohort study. \u003cem\u003eJ. Immunother Cancer\u003c/em\u003e. \u003cb\u003e8\u003c/b\u003e \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1136/jitc-2019-000381\u003c/span\u003e\u003cspan address=\"10.1136/jitc-2019-000381\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChowell, D. et al. Improved prediction of immune checkpoint blockade efficacy across multiple cancer types. \u003cem\u003eNat. Biotechnol.\u003c/em\u003e \u003cb\u003e40\u003c/b\u003e, 499\u0026ndash;506. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41587-021-01070-8\u003c/span\u003e\u003cspan address=\"10.1038/s41587-021-01070-8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChu, Y. et al. Pan-cancer T cell atlas links a cellular stress response state to immunotherapy resistance. \u003cem\u003eNat. Med.\u003c/em\u003e \u003cb\u003e29\u003c/b\u003e, 1550\u0026ndash;1562. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41591-023-02371-y\u003c/span\u003e\u003cspan address=\"10.1038/s41591-023-02371-y\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZheng, L. et al. Pan-cancer single-cell landscape of tumor-infiltrating T cells. \u003cem\u003eScience\u003c/em\u003e \u003cb\u003e374\u003c/b\u003e, abe6474. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1126/science.abe6474\u003c/span\u003e\u003cspan address=\"10.1126/science.abe6474\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYost, K. E. et al. Clonal replacement of tumor-specific T cells following PD-1 blockade. \u003cem\u003eNat. Med.\u003c/em\u003e \u003cb\u003e25\u003c/b\u003e, 1251\u0026ndash;1259. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41591-019-0522-3\u003c/span\u003e\u003cspan address=\"10.1038/s41591-019-0522-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTay, C., Tanaka, A. \u0026amp; Sakaguchi, S. Tumor-infiltrating regulatory T cells as targets of cancer immunotherapy. \u003cem\u003eCancer Cell.\u003c/em\u003e \u003cb\u003e41\u003c/b\u003e, 450\u0026ndash;465. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.ccell.2023.02.014\u003c/span\u003e\u003cspan address=\"10.1016/j.ccell.2023.02.014\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFang, X., Bogomolovas, J., Trexler, C. \u0026amp; Chen, J. The BAG3-dependent and -independent roles of cardiac small heat shock proteins. \u003cem\u003eJCI Insight\u003c/em\u003e. \u003cb\u003e4\u003c/b\u003e \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1172/jci.insight.126464\u003c/span\u003e\u003cspan address=\"10.1172/jci.insight.126464\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYang, Y. et al. Pan-cancer single-cell dissection reveals phenotypically distinct B cell subtypes. \u003cem\u003eCell\u003c/em\u003e 187, 4790\u0026ndash;4811 e4722, (2024). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.cell.2024.06.038\u003c/span\u003e\u003cspan address=\"10.1016/j.cell.2024.06.038\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eThommen, D. S. et al. A transcriptionally and functionally distinct PD-1(+) CD8(+) T cell pool with predictive potential in non-small-cell lung cancer treated with PD-1 blockade. \u003cem\u003eNat. Med.\u003c/em\u003e \u003cb\u003e24\u003c/b\u003e, 994\u0026ndash;1004. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41591-018-0057-z\u003c/span\u003e\u003cspan address=\"10.1038/s41591-018-0057-z\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMietz, J. et al. Human effector CD8(+) T cells with an activated and exhausted-like phenotype control tumour growth in vivo in a humanized tumour model. \u003cem\u003eEBioMedicine\u003c/em\u003e \u003cb\u003e106\u003c/b\u003e, 105240. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.ebiom.2024.105240\u003c/span\u003e\u003cspan address=\"10.1016/j.ebiom.2024.105240\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOliveira, G. et al. Preexisting tumor-resident T cells with cytotoxic potential associate with response to neoadjuvant anti-PD-1 in head and neck cancer. \u003cem\u003eSci. Immunol.\u003c/em\u003e \u003cb\u003e8\u003c/b\u003e, eadf4968. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1126/sciimmunol.adf4968\u003c/span\u003e\u003cspan address=\"10.1126/sciimmunol.adf4968\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKumagai, S. et al. The PD-1 expression balance between effector and regulatory T cells predicts the clinical efficacy of PD-1 blockade therapies. \u003cem\u003eNat. Immunol.\u003c/em\u003e \u003cb\u003e21\u003c/b\u003e, 1346\u0026ndash;1358. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41590-020-0769-3\u003c/span\u003e\u003cspan address=\"10.1038/s41590-020-0769-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDeTomaso, D. \u0026amp; Yosef, N. Hotspot identifies informative gene modules across modalities of single-cell genomics. \u003cem\u003eCell Syst\u003c/em\u003e 12, 446\u0026ndash;456 e449, (2021). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.cels.2021.04.005\u003c/span\u003e\u003cspan address=\"10.1016/j.cels.2021.04.005\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang, L. et al. Single-Cell Analyses Inform Mechanisms of Myeloid-Targeted Therapies in Colon Cancer. \u003cem\u003eCell\u003c/em\u003e 181, 442\u0026ndash;459 e429, (2020). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.cell.2020.03.048\u003c/span\u003e\u003cspan address=\"10.1016/j.cell.2020.03.048\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMantovani, A., Allavena, P., Marchesi, F. \u0026amp; Garlanda, C. Macrophages as tools and targets in cancer therapy. \u003cem\u003eNat. Rev. Drug Discov\u003c/em\u003e. \u003cb\u003e21\u003c/b\u003e, 799\u0026ndash;820. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41573-022-00520-5\u003c/span\u003e\u003cspan address=\"10.1038/s41573-022-00520-5\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNalio Ramos, R. et al. Tissue-resident FOLR2(+) macrophages associate with CD8(+) T cell infiltration in human breast cancer. \u003cem\u003eCell\u003c/em\u003e 185, 1189\u0026ndash;1207 e1125, (2022). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.cell.2022.02.021\u003c/span\u003e\u003cspan address=\"10.1016/j.cell.2022.02.021\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLavin, Y. et al. Innate Immune Landscape in Early Lung Adenocarcinoma by Paired Single-Cell Analyses. \u003cem\u003eCell\u003c/em\u003e 169, 750\u0026ndash;765 e717, (2017). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.cell.2017.04.014\u003c/span\u003e\u003cspan address=\"10.1016/j.cell.2017.04.014\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHouse, I. G. et al. Macrophage-Derived CXCL9 and CXCL10 Are Required for Antitumor Immune Responses Following Immune Checkpoint Blockade. \u003cem\u003eClin. Cancer Res.\u003c/em\u003e \u003cb\u003e26\u003c/b\u003e, 487\u0026ndash;504. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1158/1078-0432.CCR-19-1868\u003c/span\u003e\u003cspan address=\"10.1158/1078-0432.CCR-19-1868\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePeperzak, V. et al. CD8\u0026thinsp;+\u0026thinsp;T cells produce the chemokine CXCL10 in response to CD27/CD70 costimulation to promote generation of the CD8\u0026thinsp;+\u0026thinsp;effector T cell pool. \u003cem\u003eJ. Immunol.\u003c/em\u003e \u003cb\u003e191\u003c/b\u003e, 3025\u0026ndash;3036. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.4049/jimmunol.1202222\u003c/span\u003e\u003cspan address=\"10.4049/jimmunol.1202222\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCheng, S. et al. A pan-cancer single-cell transcriptional atlas of tumor infiltrating myeloid cells. \u003cem\u003eCell\u003c/em\u003e 184, 792\u0026ndash;809 e723, (2021). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.cell.2021.01.010\u003c/span\u003e\u003cspan address=\"10.1016/j.cell.2021.01.010\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSharma, P. et al. Immune checkpoint therapy-current perspectives and future directions. \u003cem\u003eCell\u003c/em\u003e \u003cb\u003e186\u003c/b\u003e, 1652\u0026ndash;1669. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.cell.2023.03.006\u003c/span\u003e\u003cspan address=\"10.1016/j.cell.2023.03.006\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSt Paul, M. \u0026amp; Ohashi, P. S. The Roles of CD8(+) T Cell Subsets in Antitumor Immunity. \u003cem\u003eTrends Cell. Biol.\u003c/em\u003e \u003cb\u003e30\u003c/b\u003e, 695\u0026ndash;704. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.tcb.2020.06.003\u003c/span\u003e\u003cspan address=\"10.1016/j.tcb.2020.06.003\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLegler, D. F. et al. B cell-attracting chemokine 1, a human CXC chemokine expressed in lymphoid tissues, selectively attracts B lymphocytes via BLR1/CXCR5. \u003cem\u003eJ. Exp. Med.\u003c/em\u003e \u003cb\u003e187\u003c/b\u003e, 655\u0026ndash;660. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1084/jem.187.4.655\u003c/span\u003e\u003cspan address=\"10.1084/jem.187.4.655\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (1998).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGunn, M. D. et al. A B-cell-homing chemokine made in lymphoid follicles activates Burkitt's lymphoma receptor-1. \u003cem\u003eNature\u003c/em\u003e \u003cb\u003e391\u003c/b\u003e, 799\u0026ndash;803. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/35876\u003c/span\u003e\u003cspan address=\"10.1038/35876\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (1998).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXu, T. et al. Identification of CXCL13 as a potential biomarker in clear cell renal cell carcinoma via comprehensive bioinformatics analysis. \u003cem\u003eBiomed. Pharmacother\u003c/em\u003e. \u003cb\u003e118\u003c/b\u003e, 109264. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.biopha.2019.109264\u003c/span\u003e\u003cspan address=\"10.1016/j.biopha.2019.109264\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGroeneveld, C. S. et al. Tertiary lymphoid structures marker CXCL13 is associated with better survival for patients with advanced-stage bladder cancer treated with immunotherapy. \u003cem\u003eEur. J. Cancer\u003c/em\u003e. \u003cb\u003e148\u003c/b\u003e, 181\u0026ndash;189. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.ejca.2021.01.036\u003c/span\u003e\u003cspan address=\"10.1016/j.ejca.2021.01.036\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYang, M. et al. CXCL13 shapes immunoactive tumor microenvironment and enhances the efficacy of PD-1 checkpoint blockade in high-grade serous ovarian cancer. \u003cem\u003eJ. Immunother Cancer\u003c/em\u003e. \u003cb\u003e9\u003c/b\u003e \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1136/jitc-2020-001136\u003c/span\u003e\u003cspan address=\"10.1136/jitc-2020-001136\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGoswami, S. et al. ARID1A mutation plus CXCL13 expression act as combinatorial biomarkers to predict responses to immune checkpoint therapy in mUCC. \u003cem\u003eSci. Transl Med.\u003c/em\u003e \u003cb\u003e12\u003c/b\u003e \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1126/scitranslmed.abc4220\u003c/span\u003e\u003cspan address=\"10.1126/scitranslmed.abc4220\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBassez, A. et al. A single-cell map of intratumoral changes during anti-PD1 treatment of patients with breast cancer. \u003cem\u003eNat. Med.\u003c/em\u003e \u003cb\u003e27\u003c/b\u003e, 820\u0026ndash;832. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41591-021-01323-8\u003c/span\u003e\u003cspan address=\"10.1038/s41591-021-01323-8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXie, L., Ning, Z., Hua, Y., Wang, P. \u0026amp; Meng, Z. Single-cell transcriptome analysis revealed the immune profile of PD-1 blockade in gallbladder carcinoma liver metastasis. \u003cem\u003eHepatol. Commun.\u003c/em\u003e \u003cb\u003e7\u003c/b\u003e \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1097/HC9.0000000000000131\u003c/span\u003e\u003cspan address=\"10.1097/HC9.0000000000000131\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eQiu, M. Z. et al. Dynamic single-cell mapping unveils Epstein\u0026ndash;Barr virus-imprinted T-cell exhaustion and on-treatment response. \u003cem\u003eSignal. Transduct. Target. Ther.\u003c/em\u003e \u003cb\u003e8\u003c/b\u003e, 370. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41392-023-01622-1\u003c/span\u003e\u003cspan address=\"10.1038/s41392-023-01622-1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMueller, A. et al. Transcriptional and Clonal Characterization of Cytotoxic T Cells in Crescentic Glomerulonephritis. \u003cem\u003eJ. Am. Soc. Nephrol.\u003c/em\u003e \u003cb\u003e34\u003c/b\u003e, 1003\u0026ndash;1018. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1681/ASN.0000000000000116\u003c/span\u003e\u003cspan address=\"10.1681/ASN.0000000000000116\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang, Y. et al. Targeting CD96 overcomes PD-1 blockade resistance by enhancing CD8\u0026thinsp;+\u0026thinsp;TIL function in cervical cancer. \u003cem\u003eJ. Immunother Cancer\u003c/em\u003e. \u003cb\u003e10\u003c/b\u003e \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1136/jitc-2021-003667\u003c/span\u003e\u003cspan address=\"10.1136/jitc-2021-003667\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang, X. Q. et al. Spatial predictors of immunotherapy response in triple-negative breast cancer. \u003cem\u003eNature\u003c/em\u003e \u003cb\u003e621\u003c/b\u003e, 868\u0026ndash;876. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41586-023-06498-3\u003c/span\u003e\u003cspan address=\"10.1038/s41586-023-06498-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePolanski, K. et al. BBKNN: fast batch alignment of single cell transcriptomes. \u003cem\u003eBioinformatics\u003c/em\u003e \u003cb\u003e36\u003c/b\u003e, 964\u0026ndash;965. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1093/bioinformatics/btz625\u003c/span\u003e\u003cspan address=\"10.1093/bioinformatics/btz625\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTraag, V. A., Waltman, L. \u0026amp; van Eck, N. J. From Louvain to Leiden: guaranteeing well-connected communities. \u003cem\u003eSci. Rep.\u003c/em\u003e \u003cb\u003e9\u003c/b\u003e, 5233. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41598-019-41695-z\u003c/span\u003e\u003cspan address=\"10.1038/s41598-019-41695-z\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGuo, X. et al. Global characterization of T cells in non-small-cell lung cancer by single-cell sequencing. \u003cem\u003eNat. Med.\u003c/em\u003e \u003cb\u003e24\u003c/b\u003e, 978\u0026ndash;985. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41591-018-0045-3\u003c/span\u003e\u003cspan address=\"10.1038/s41591-018-0045-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang, L. et al. Lineage tracking reveals dynamic relationships of T cells in colorectal cancer. \u003cem\u003eNature\u003c/em\u003e \u003cb\u003e564\u003c/b\u003e, 268\u0026ndash;272. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41586-018-0694-x\u003c/span\u003e\u003cspan address=\"10.1038/s41586-018-0694-x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCao, J. et al. The single-cell transcriptional landscape of mammalian organogenesis. \u003cem\u003eNature\u003c/em\u003e \u003cb\u003e566\u003c/b\u003e, 496\u0026ndash;502. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41586-019-0969-x\u003c/span\u003e\u003cspan address=\"10.1038/s41586-019-0969-x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiu, Z. et al. Progenitor-like exhausted SPRY1(+)CD8(+) T cells potentiate responsiveness to neoadjuvant PD-1 blockade in esophageal squamous cell carcinoma. \u003cem\u003eCancer Cell.\u003c/em\u003e \u003cb\u003e41\u003c/b\u003e (e1859), 1852\u0026ndash;1870. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.ccell.2023.09.011\u003c/span\u003e\u003cspan address=\"10.1016/j.ccell.2023.09.011\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBi, K. et al. Tumor and immune reprogramming during immunotherapy in advanced renal cell carcinoma. \u003cem\u003eCancer Cell\u003c/em\u003e 39, 649\u0026ndash;661 e645, (2021). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.ccell.2021.02.015\u003c/span\u003e\u003cspan address=\"10.1016/j.ccell.2021.02.015\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJin, S. et al. Inference and analysis of cell-cell communication using CellChat. \u003cem\u003eNat. Commun.\u003c/em\u003e \u003cb\u003e12\u003c/b\u003e, 1088. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41467-021-21246-9\u003c/span\u003e\u003cspan address=\"10.1038/s41467-021-21246-9\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2021).\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"CD8 + T effector cells, Immune checkpoint blockade, Machine learning, Pan cancer, PD-1","lastPublishedDoi":"10.21203/rs.3.rs-6317412/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6317412/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eProgrammed cell death protein-1 (PD-1) blockade therapy have shown significant efficacy in cancer immunotherapy. However, low response rates and individual variability remain challenges. Currently, a universal biomarker to assess immunotherapy efficacy across various cancer types is lacking. In this study, single-cell RNA sequencing was applied to samples from seven cancer types, alongside bulk RNA-seq data from eight additional cancer types. LASSO regression and 15 machine learning algorithms were employed to construct 152 predictive models for immunotherapy efficacy. The results indicated that CD8\u0026thinsp;+\u0026thinsp;effector T cells (CD8_Teff) in responders exhibited high infiltration and an activated, exhaustion-like phenotype. A predictive model based on seven effector T cell immunotherapy response genes (ETIRGS) effectively distinguished between responders and non-responders. The high-predicted scoring group exhibited significantly higher infiltration of CD8\u0026thinsp;+\u0026thinsp;T cells and M1 macrophages than the low-predicted scoring group, along with elevated stromal and immune scores. Macrophages in responders acquired a pro-inflammatory phenotype upon activation by CD8_Teff cells, thereby enhancing the immune response. This study provides potential cross-cancer predictive biomarkers for PD-1 blockade therapy.\u003c/p\u003e","manuscriptTitle":"Constructing a Predictive Model for PD-1 Blockade Therapy in Pan-Cancer Based on Machine Learning","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-04-03 08:27:38","doi":"10.21203/rs.3.rs-6317412/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"de93a6b2-d16e-4cfe-bce6-e121c034c950","owner":[],"postedDate":"April 3rd, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":46441373,"name":"Health sciences/Biomarkers"},{"id":46441374,"name":"Health sciences/Oncology"}],"tags":[],"updatedAt":"2025-10-03T02:38:33+00:00","versionOfRecord":[],"versionCreatedAt":"2025-04-03 08:27:38","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6317412","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6317412","identity":"rs-6317412","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}