Abstract
The effects of methylation/autophagy-related genes (MARGs) and immune infiltration in the tumor microenvironment on the prognosis of laryngeal cancer were comprehensively explored in this study. Survival analysis screened out 126 MARGs and 10 immune cells potentially associated with the prognosis of laryngeal carcinoma. Cox and lasso regression analyses were then used to select 8 MARGs (CAPN10, DAPK2, MBTPS2, ST13, CFLAR, FADD, PEX14 and TSC2) and 2 immune cells (Eosinophil and Mast cell) to obtain the prognostic risk scoring system (pRS). The pRS was used to establish a risk prediction model for the prognosis of laryngeal cancer. The predictive ability of the prediction model was evaluated by GEO datasets and our clinical samples. Further analysis revealed that pRS is highly associated with single nucleotide polymorphism (SNP), copy number variation (CNV), immune checkpoint blockade (ICB) therapy and tumor microenvironment. Moreover, the screened pRS-related ceRNA network and circ_0002951/miR-548k/HAS2 pathway provide potential therapeutic targets and biomarkers of laryngocarcinoma. Based on the clustering results of pRS-related genes, single cells were then genotyped and revealed by integrated scRNA-seq in laryngeal cancer samples. Fibroblasts were found enriched in high risk cell clusters at the scRNA-seq level. Fibroblast-related ligand-receptor interactions were then exposed and a neural network-based deep learning model based on these pRS-related hub gene signatures was also established with a high accuracy in cell type prediction. In conclusion, the combination of single-cell and transcriptome laryngeal carcinoma landscape analyses can investigate the link between the tumor microenvironmental and prognostic characteristics.