Machine learning-based prognostic model and single-cell transcriptomic integration for identifying brain metastasis-associated malignant subpopulations and potential therapeutic targets in lung adenocarcinoma.
A comprehensive single-cell atlas of the BrM TME is unveiled and a robust 12-gene BrM-informed prognostic signature for LUAD is established, validated across multiple LUAD survival cohorts.
Abstract
Background
Lung adenocarcinoma (LUAD) is a highly prevalent and lethal form of lung cancer. Brain metastasis (BrM) is a major cause of mortality in patients with LUAD. The tumor microenvironment (TME) of LUAD BrM remains incompletely characterized, and there is an urgent need to identify biologically relevant biomarkers and potential therapeutic vulnerabilities.
Methods
Single-cell transcriptomic data from LUAD BrM samples were processed through dimensionality reduction, clustering, and cell annotation. BrM-associated core genes were identified from BrM scRNA-seq datasets using inferCNV and hdWGCNA. A prognostic model for LUAD was then constructed using 101 combinatorial approaches derived from 10 machine learning algorithms and validated in LUAD cohorts. We further performed supportive analyses including enrichment, pseudotime trajectory, cell-cell communication, immune infiltration, and single-cell-based drug sensitivity prediction to provide additional biological context. Key findings were validated by immunohistochemistry (IHC) in human tissues and by qRT-PCR and IHC in a mouse BrM model. Functional assays were performed in PC9-BrM cells to assess the roles of SEC61G and CFL1 in LUAD BrM.
Results
Our study unveiled a comprehensive single-cell atlas of the BrM TME and established a robust 12-gene BrM-informed prognostic signature for LUAD, validated across multiple LUAD survival cohorts. Functional analysis revealed dysregulated cell cycle control in BrM progression. Immune profiling demonstrated significantly reduced CD4+ and CD8+ T-cell infiltration in high-risk subgroups. Cell-cell interaction analysis highlighted the MIF-CD74/CXCR4 and MIF-CD74/CD44 signaling axes as key regulators of TME remodeling. IHC in human tissues and validation using a mouse BrM model supported the biological relevance of SEC61G and CFL1 in LUAD BrM. qRT-PCR further showed elevated expression of SEC61G and CFL1 mRNA in brain-metastatic cells. Knockdown of SEC61G or CFL1 significantly inhibited PC9-BrM cell viability, proliferation, and migration.
Conclusions
This study identified genes associated with LUAD BrM and established a machine-learning-based prognostic signature for LUAD. Further experimental validation supports SEC61G and CFL1 as candidate biomarkers with potential therapeutic relevance.
Background: Prostate cancer (PCa) is a leading cause of cancer-related mortality worldwide, highlighting the need for improved prognostic tools. The integration of artificial intelligence (AI) and machine learning (ML) with multi-omics data offers new opportunities for biomarker discovery and risk stratification. Methods: We integrated bulk transcriptomic data from GSE116918 (training, n = 248) and three cross-cohort consistency evaluation cohorts (TCGA-PRAD, GSE70769, GSE46602), focusing on 1087 epithelial–mesenchymal transition (EMT)-associated genes. Using consensus clustering, weighted gene co-expression network analysis (WGCNA), and 91 machine learning algorithm combinations (including Random Forest, Lasso, and CoxBoost), we constructed a prognostic signature. SHAP analysis was used for model interpretability. Single-cell RNA sequencing (scRNA-seq, GSE268307, 10,672 cells) and spatial transcriptomics (10× Genomics Visium FFPE) provided hypothesis-generating evidence; spatial analysis was based on one tissue section. Results: A three-gene signature (INHBA, FAP, ITGBL1) effectively stratified patients into high- and low-risk groups, with the high-risk group showing significantly worse metastasis-free survival (HR = 1.61, 95% CI: 1.39–1.87; 4-year AUC = 0.93 in the training cohort; external AUCs ranged from 0.62 to 0.77). CytoTRACE inferred high differentiation potential of COMP+ fibroblasts, and Monocle3 inferred a transcriptional transition from COMP+ toward NELL2+ fibroblasts. BayesPrism deconvolution suggested that high inferred COMP+ fibroblast abundance was associated with poor prognosis and advanced T stage. NicheNet analysis prioritized BMP7 as a key upstream ligand, with downstream targets enriched in TGF-β signaling and stem cell pluripotency pathways. Conclusions: This study presents a machine learning-based multi-omics framework for prostate cancer risk stratification. The three-gene signature provides a new exploratory prognostic model while inferring a COMP+ to NELL2+ transcriptional transition. These findings may inform future hypothesis-driven studies of treatment sensitivity, pending experimental validation, and demonstrate the value of AI-driven multi-omics integration for precision oncology.
Xueqian Zhang, Wei Zhang, Zheng Wang et al.· Genes· 0 citations
Background Epstein–Barr virus–associated gastric cancer (EBVaGC) is a distinct molecular subtype of gastric cancer, but reliable biomarkers linking EBV‐related biology, prognosis, and microenvironmental remodeling remain limited. Methods Differentially expressed genes associated with EBVaGC were identified from TCGA and GEO datasets. Weighted gene co‐expression network analysis and machine learning were used to screen core genes. Functional enrichment, diagnostic and prognostic modeling, immune infiltration analysis, single‐cell transcriptomic analysis of a public scRNA‐seq dataset, cell–cell communication analysis, drug sensitivity prediction, external cohort validation, and RT‐qPCR validation in gastric cancer cell lines were subsequently performed. Results Five core genes (ASPA, CHODL, GNG7, P2RY14, and PI16) were identified. The combined classifier showed high apparent diagnostic performance in the discovery datasets (AUC = 1.000 in the EBV cohort and 0.981 in TCGA–STAD) and retained value in GSE27342 (AUC = 0.771). RT‐qPCR confirmed differential expression of these genes between a gastric epithelial cell line and a gastric cancer cell line, providing general tumor‐versus‐normal experimental support rather than EBV‐specific validation. A five‐gene risk signature stratified overall survival, with 1‐, 2‐, and 3‐year AUCs of 0.634, 0.645, and 0.622, which improved to 0.716, 0.768, and 0.693 after integration with age and stage. Single‐cell analysis localized the strongest signature signal to fibroblasts and B cells and revealed enhanced MIF‐, PTN‐, and TNFSF13B‐related communication in tumors. High‐risk tumors showed higher predicted IC50 values for paclitaxel and 5‐fluorouracil. Conclusion We identified a five‐gene signature with strong diagnostic value and biologically meaningful prognostic relevance in EBVaGC, highlighting fibroblast‐ and B‐cell‐associated microenvironmental programs in aggressive disease.
Luyu Jin, Man Jiang, Yulong Pan et al.· Genetics Research· 0 citations
Glioblastoma (GBM) is the most common primary intracranial malignancy in adults, characterized by poor survival and high mortality. Emerging evidence suggests that macrophage-associated programmed cell death (MacPCD) plays a critical role in GBM pathogenesis. However, the underlying mechanisms remain poorly understood. This study aimed to identify MacPCD-related prognostic genes in GBM and explore their functional roles.
Transcriptomic data from the GSE68848 dataset were integrated with Macrophage-associated programmed cell death-related genes (MacPCD-RGs) to identify differentially expressed genes (DEGs). Univariate Cox and LASSO regression analyses were performed using the TCGA-GBM training set to construct a prognostic risk model. Beyond prognostic stratification, we conducted a comprehensive multi-omic landscape analysis, including gene set enrichment analysis (GSEA), tumor microenvironment (TME) characterization, tumor mutational burden (TMB) assessment, immunotherapy response prediction, and drug sensitivity prediction. Finally, single-cell RNA sequencing (scRNA-seq) was employed to resolve microenvironmental heterogeneity, identify key cell types and elucidate intercellular communication and developmental trajectories.
Analysis of GSE68848 identified 902 DEGs, of which five intersected with MacPCD-RGs.
FN1
and
TIMP1
were subsequently identified as core prognostic markers. The risk model demonstrated superior predictive performance across the CGGA-325 and GSE83300 validation cohorts. Functional analysis linked the risk score to specific signaling pathways,
PTEN
mutations, infiltration of immune cell subsets (e.g. NKT cells) and sensitivity to Trametinib. Tumor-associated macrophages (TAMs) were identified as the key cell type, exhibiting intense interaction with pericytes and enrichment in fructose/mannose metabolism. Furthermore, pseudotime analysis revealed that
FN1
and
TIMP1
expression peaked during the initial stages of TAM differentiation.
This study identified
FN1
and
TIMP1
as pivotal MacPCD-related prognostic genes in GBM. The risk model based on these markers exhibits moderate predictive performance, offering a reliable tool for clinical prognosis and paving the way for personalized immunotherapy strategies.
Single‐cell transcriptomics and machine learning methods are increasingly used to identify immune‐related biomarkers in solid tumors, yet their combined application to microenvironment‐related drivers of therapeutic resistance in clear cell renal cell carcinoma (ccRCC) is still limited. Here, we investigated the biological and clinical significance of myeloid‐derived growth factor (MYDGF) through an integrative strategy spanning single‐cell profiling, bulk multiomics, and functional validation. Analysis of scRNA‐seq data (GSE156632) revealed that MYDGF is preferentially detected in malignant epithelial subpopulations and associated with the composition of myeloid and lymphoid compartments. Integration with TCGA‐KIRC transcriptomic and clinical datasets demonstrated strong associations between MYDGF expression and immune‐checkpoint activation, immune dysfunction signatures, and PI3K/AKT–MAPK pathway activity. Tumors with high MYDGF expression exhibited an immune‐infiltrated yet functionally impaired microenvironment and were predicted to show reduced responsiveness to immune checkpoint blockade. Differential expression and enrichment analyses further highlighted MYDGF‐associated genes involved in inflammatory, extracellular, and receptor‐binding functions. A machine learning pipeline using LASSO Cox regression identified a preliminary 19‐gene MYDGF‐related prognostic gene set that requires further validation. Functional experiments confirmed that MYDGF knockdown suppressed proliferation, migration, and invasion in ccRCC cells. Overall, our analyses characterize MYDGF as a microenvironment‐related biomarker linked to immune‐associated features, signaling‐associated alterations, and adverse prognosis in ccRCC. These results nominate MYDGF as a candidate prognostic biomarker and show the value of pairing single‐cell resolution with computational modeling for biomarker discovery in renal cancer.
Yingkun Xu, Guandu Li, Xinxiu Ren et al.· Human Mutation· 0 citations
INTRODUCTION
Cytokines play an important role in modulating the tumor microenvironment (TME) in glioblastoma multiforme (GBM). However, little work has focused on developing a prognostic model for GBM using cytokine signatures.
METHODS
Herein, we combined several GBM datasets, such as GSE163120, TCGA-GBM, CGGA-693, CGGA-325, GSE16011, the Rembrandt dataset, and GTEx. We combined the characteristic genes of different cell types in GBM samples with cytokine-associated genes identified in our previous research to identify single-cell cytokine-related genes (scCRGs). The enrichment scores of scCRGs in each cell were computed, and cells were stratified into a high-expression group and a low-expression group based on the median enrichment score. Differentially expressed scCRGs in the high-expression group were identified using the "limma" R package. Then, 117 combinations of machine learning (ML) algorithms were used to construct the new cytokine-related signature (CRS) prediction model. Moreover, we used consensus clustering algorithms to perform novel clustering analyses on GBM and subsequently conducted comprehensive immune profiling, response-to-immune-therapy prediction, and drug-sensitivity evaluation. Finally, we validated the key molecules in the model using tissue microarrays.
RESULTS
We identified 17 scCRGs that are mostly highly expressed in microglia. The new CRS model we proposed demonstrated strong predictive performance across the three independent cohorts, outperforming existing GBM models. Based on the CRS model, we identified distinct subgroups of GBM patients who may respond favorably to immunotherapy. The key gene AEBP1 in the model is highly expressed in GBM tissues.
DISCUSSION
Here, we present, for the first time, a highly reproducible and patient-specific prognostic prediction model based on a cytokine-related gene signature, combining several ML techniques and a large set of bioinformatic features. Compared with other published models, the proposed model is more stable and produces better predictions. Based on the model, we could identify separate subgroups of GBM patients who might respond better to immunotherapy, providing actionable information for the development of precision medicine approaches to treating GBM.
CONCLUSION
CRS has the potential to be an effective and promising strategy for improving clinical outcomes in patients with GBM.
Longxiao Zhang, Xinyang Yan, Yunfei Zhou et al.· Current Cancer Drug Targets· 0 citations
It is suggested that TRIP6 may be involved in linking microenvironmental signaling to invasive phenotypes, with potential value for prognostic stratification and therapeutic exploration.
Qiwei Cui, Xiao-Hong Gao, Xiaoqing Wu et al.· Frontiers in Cell and Develo...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
