Construction Of An Efficient Prognostic Model For Lung Adenocarcinoma Metastasis Based On Single Cell RNA Transcriptome Data

Lung adenocarcinoma is characterised by its tendency to metastasise and is the subtype of lung cancer with the highest mortality rate.The metastasis of tumour cells is a complex system that requires the use of blood circulation and lymphatic vessels,involves multiple cell types,microenvironments and tissues,is poor prognosis and systemic disease,which is a major contributor to the high mortality rate of cancer.The rise of high-throughput sequencing technologies in recent years has made it possible to obtain information on individual cells.Single-cell RNA transcriptome data can reveal cellular heterogeneity in the metastatic microenvironment of different organs,providing technical support for the study of cancer metastasis.Therefore,this paper investigates the prognosis of lung adenocarcinoma metastasis based on single-cell RNA transcriptome data.Firstly,single cell RNA transcriptome data containing normal tissue,primary tissue,brain metastasis,bone metastasis and lymph node metastasis tissue of lung adenocarcinoma were downloaded from the GEO database,and the high quality cells were retained through quality control and dimensional reduction clustering of the data.On this basis,nine cell types were identified using both typical marker genes and Single R automated annotation,identifying 25,099 epithelial cells,23,546 T cells.Secondly,the 10 T cell subpopulations of CD8+ and CD4+ were identified by reclustering among the defined T cells based on surface antigens and function.The metastatic microenvironment of T cell subsets was studied at the primary and metastatic sites,and the immune response was found to be stronger at the metastatic site than at the primary site.During metastasis,CD4 na(?)ve T cells underwent differentiation and CD4 regulatory T cells suppressed immune activity.Malignant epithelial cells were inferred from the characteristics of copy number variation in defined epithelial cells and were found to deteriorate more at the metastatic site than at the primary site.Gene differential expression analysis was then used at the malignant cell level to screen for metastasisrelated genes as candidates for the construction of prognostic models.Finally,9 key prognostic genes were trained and validated using machine learning and other statistical methods in combination with 407 cases of bulk sequenced lung adenocarcinoma data with clinical information downloaded from the TCGA database.In addition,clinical indicators with strong correlation with prognostic genes,such as T-stage,N-stage and pathologic stage were screened to construct a nomogram model to predict metastatic survival in patients with lung adenocarcinoma in a comprehensive clinical setting.Validated using ROC curves,calibration curves and survival analysis,the model was found to have good predictive accuracy for the overall survival of patients at 1 and 3years.Through somatic mutation analysis and drug sensitivity analysis,it can be found that the population with a high nomogram score has a poor prognosis,accompanied by a high tumor somatic mutation burden,and is more sensitive to chemotherapy drugs and targeted therapy drugs.By performing enrichment analysis in the high and low nomogram score groups,it was found that the occurrence of these conditions may be related to the upregulation of pathways such as cell cycle,ECM receptor interactions,the P53 signalling pathway,and spliceosome pathway.