Harnessing Anoikis-Related Gene Signatures for Immunotherapy Precision: A Machine Learning-Driven Predictive and...

Purpose: Cancer remains a leading cause of death, with immune checkpoint inhibitors (ICIs) offering promising but heterogeneous therapeutic outcomes. This study investigates the role of anoikis-related genes (ARGs) in immunotherapy efficacy and develops a machine learning-based predictive model for immunotherapy response. Methods: We integrated single-cell RNA sequencing (scRNA-seq) and multiomics data from various cancer types to identify ARGs associated with immunotherapy response. A total of 41 anoikis-related differential genes (Anoikis.Sig) were identified through correlation and enrichment analyses. To assess the predictive value of these ARGs, we constructed machine learning models using seven classifiers, including Naive Bayes (NB), Random Forest (RF), Support Vector Machine (SVM), and others. We also validated the model across multiple independent data sets. Additionally, we conducted CRISPR-based screening to identify immune-resistant ARGs that influence tumor immunity. Prognostic analysis was performed on a pan-cancer data set to evaluate the clinical relevance of the identified ARGs. Results: Our analyses established a significant association between Anoikis.Sig and immunotherapy response. Specifically, lower Anoikis Scores were strongly correlated with improved immunotherapy outcomes in melanoma and BCC cohorts. Among seven classifiers, the SVM-based model achieved superior predictive performance with an AUC of 0.782, consistently outperforming established gene signatures across multiple validation data sets. Furthermore, integrated CRISPR screening pinpointed key immune-resistant genes, such as BCL2L1, ITGAV, and PTK2, as potential drivers of tumor progression. The 11-gene Key-Anoikis.Sig not only demonstrated robust prognostic value across 30 cancer types but also showed a remarkably strong survival association in hepatocellular carcinoma (HCC). Conclusion: ARGs are essential modulators of the tumor-immune landscape and serve as high-performance biomarkers for predicting ICI efficacy. The machine learning-driven model developed in this study provides a reliable tool for precise patient stratification and personalized therapeutic guidance. Our findings underscore the potential of targeting ARGs as a novel strategy to enhance immunotherapy outcomes.