郭宇琪 1蔡雨思 1王浩 1郭玥 1王春 1陈嘉康1
作者信息
- 1. 北京大学深圳医院病理科,深圳 518000
- 折叠
摘要
Abstract
Objective:The high recurrence and drug resistance of lung adenocarcinoma(LUAD)are closely associated with the heterogeneity of its tumor microenvironment(TME).Recent studies have revealed that intratumoral microbiota,particularly members of the phylum Firmicutes,can modulate host gene expression and immune responses through metabolites such as butyrate,thereby influencing tumor progression.However,the molecular characteristics of Firmicutes-host interactions and their prognostic implications in LUAD remain unclear.This study aims to identify Firmicutes-associated host gene signatures in LUAD and to construct a prognostic model integrating microbiota-host interaction features to guide precision risk stratification and targeted therapy.
Methods:Firmicutes-related genes significantly associated with LUAD were screened from the Bacteria in Cancer(BIC)database,and transcriptomic and clinical data were obtained from The Cancer Genome Atlas(TCGA).10 combinations of machine learning algorithms were employed to construct prognostic models,which were validated using Kaplan-Meier survival and receiver operating characteristic(ROC)analyses.The final model was optimized by least absolute shrinkage and selection operator(LASSO)-Cox regression,and risk scores were calculated.Univariate and multivariate Cox regression analyses were performed to evaluate the prognostic value of the risk score.Gene Ontology(GO)and Kyoto Encyclopedia of Genes and Genomes(KEGG)analyses were conducted to explore biological pathways associated with high-and low-risk groups,and immune infiltration patterns were assessed to investigate potential mechanisms.
Results:LUAD patients were stratified into two Firmicutes-related molecular subtypes with significantly different prognosis(P=0.01).Among the 10 machine learning model combinations,the StepCox[forward]+gradient boosting machines(GBM)model achieved the best performance,with concordance indices(C-index)of 0.78 in the training set and 0.64 in the testing set.Based on calculated risk scores,patients were divided into high-and low-risk groups.Kaplan-Meier analysis demonstrated that high-risk patients had significantly shorter overall survival than low-risk patients[training set:hazard ratio(HR)=5.579,95%confidence interval(CI)3.925 to 7.930,P<0.001;testing set:HR=1.991,95%CI 1.170 to 3.387,P=0.011].ROC analysis indicated that the model achieved good predictive performance,with area under the curve(AUC)values for 1-,2-,3-,and 5-year survival of 0.772 and 0.718,0.815 and 0.739,0.830 and 0.596,and 0.851 and 0.530 in the training and testing sets,respectively.After LASSO-Cox optimization,an eight-gene prognostic model was established.Both univariate and multivariate Cox analyses confirmed that the risk score was an independent prognostic factor positively correlated with mortality risk(P<0.05).GO and KEGG analyses revealed significant enrichment of cell cycle-related pathways in different risk groups(P<0.05).Immune infiltration analysis showed that resting memory CD4+T cells,naive B cells,plasma cells,resting mast cells,and resting dendritic cells were significantly reduced in the high-risk group compared with the low-risk group(all P<0.05),suggesting that bacterial infection may influence prognosis through immune modulation.
Conclusion:The prognostic model constructed based on Firmicutes-associated host genes effectively predicts clinical outcomes in LUAD patients and offers a new strategy for individualized therapy.关键词
肺腺癌/厚壁菌门/预后模型/免疫细胞浸润/机器学习Key words
lung adenocarcinoma/Firmicutes/prognostic model/immune cell infiltration/machine learning
