AUTHOR=Chen Ziqi , Qu Lingmei , Hao Qing , Teng Shuang , Liu Shuo , Wu Qin , Yi Hongtian , Shen Xianji , Li Liang , Xu Zhaonan , Sun Yanan 

TITLE=Identification of anoikis-related genes classification patterns and immune infiltration characterization in chronic rhinosinusitis with nasal polyps based on machine learning

JOURNAL=Frontiers in Molecular Biosciences

VOLUME=Volume 12 - 2025

YEAR=2025

URL=https://www.frontiersin.org/journals/molecular-biosciences/articles/10.3389/fmolb.2025.1624300

DOI=10.3389/fmolb.2025.1624300

ISSN=2296-889X

ABSTRACT=IntroductionChronic rhinosinusitis with nasal polyps (CRSwNP) is characterized by stromal edema, albumin deposition, and pseudocyst formation. Anoikis, a process in which cells detach from the correct extracellular matrix, disrupts integrin junctions, thereby inhibiting improperly proliferating cells from growing or adhering to an inappropriate matrix. Although anoikis is implicated in immune regulation and CRSwNP pathogenesis, its specific mechanistic role remains poorly defined.MethodsThe GSE136825 and GSE179625 datasets were obtained from the GEO database and 338 anoikis-related genes (ARGs) were extracted from the literature and databases. Immune cell infiltration was analysed using the CIBERSORT algorithm. CRSwNP samples were classified via consensus clustering. Key ARGs were identified through machine learning. The diagnostic performance of candidate genes was evaluated using Receiver Operating Characteristic (ROC) analysis. Functional annotation was performed based on Gene Ontology (GO) terms, and pathway enrichment analysis was conducted using the Kyoto Encyclopedia of Genes and Genomes (KEGG) database. Regulatory networks were visualized using NetworkAnalyst and Cytoscape. Experimental validation included quantitative real-time reverse-transcription PCR (qRT-PCR), immunohistochemistry (IHC), and immunofluorescence (IF) in human tissues.ResultsConsensus clustering stratified CRSwNP patients into two distinct anoikis-related clusters. Machine learning identified four key genes: CDH3, PTHLH, PDCD4, and androgen receptor (AR). The nomogram model demonstrated high diagnostic accuracy, with an area under the receiver operating characteristic curve (AUC) >0.90. Immune infiltration analysis revealed differential immune microenvironments between clusters, with AR overexpressed in cluster 1 and PTHLH in cluster 2. Network analysis identified 862 drugs or compounds targeting AR. Experimental validation confirmed consistency between bioinformatics predictions and tissue-level expression patterns.ConclusionThis study delineates two anoikis-related molecular subtypes of CRSwNP and identifies AR and PTHLH as cluster-specific biomarkers. These findings provide novel insights for personalized therapy, drug screening, and immunomodulatory strategies in CRSwNP.