The construction and validation of ECM-related prognosis model in laryngeal squamous cell carcinoma

Abstract

BackgroundLaryngeal squamous cell carcinoma (LSCC) is a kind of common and aggressive tumor with high mortality. The application of molecular biomarkers is useful for the early diagnosis and treatment of LSCC. MethodsThe expression and clinical information were obtained from The Cancer Genome Atlas (TCGA) database. Principal components analysis (PCA) was used to discriminate between LSCC and normal samples. The hub genes were screened out through univariate and multivariate cox analyses. The Kaplan-Meier (K-M) and receiver operating characteristic (ROC) curve was used to validate the predictive performance. The single sample gene set enrichment analysis (ssGSEA), Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis were used to determine the enrichment function. Protein-Protein Interaction (PPI) network was constructed in STRING. The immune analysis was performed by ESTIMATE, IPS and xCELL. The drug sensitivity was identified with GSCA database. ResultsWe identified that 47 extracellular matrix (ECM) genes were differentially expressed in LSCC compared with normal group. Univariate and multivariate cox analysis determined that leucine-rich glioma-inactivated 4 (LGI4), matrilin 4 (MATN4), microfibrillar-associated protein 2 (MFAP2) and fibrinogen like 2 (FGL2) were closely related to the disease free survival (DSS) of LSCC. ROC curve determined that the risk model has a good predictive performance. PPI network showed the top 100 genes with high correlation of hub genes. The ssGSEA, GO and KEGG enrichment analyses determined that immune response was significantly involved in the development of LSCC. Immune infiltration analysis showed that most immune cells and immune checkpoints were inhibited in high risk score group. Drug sensitivity analysis showed that MATN4, FGL2 and LGI4 were negatively related to various drugs, while MFAP2 was positively related to many drugs. ConclusionWe established a risk model constructed with four ECM-related genes, which could effectively predict the prognosis of LSCC.