Machine learning combined with multi-omics to identify immune-related LncRNA signature as biomarkers for predicting breast cancer prognosis.

Abstract

This study developed an immune-related long non-coding RNAs (lncRNAs)-based prognostic signature by integrating multi-omics data and machine learning algorithms to predict survival and therapeutic responses in breast cancer patients. Utilizing transcriptomic and gene expression data from TCGA and GEO databases, 72 immune-related lncRNAs were identified through weighted gene co-expression network analysis (WGCNA) and ImmuLncRNA algorithms. The model was further optimized using 101 combinations of 10 machine learning approaches, ultimately constructing an immune-related lncRNA signature(IRLS) scoring system comprising nine key lncRNAs. Validated across 17 independent cohorts, the model demonstrated that high-risk patients had significantly shorter overall survival (OS) (P < 0.05), with predictive performance surpassing 95 published models (P < 0.05). Additionally, the IRLS score predicted responses to paclitaxel chemotherapy, and the low-risk group exhibited higher immune cell infiltration (P < 0.05), showing significant negative correlations with CD8A, PD-L1, tumor mutational burden (TMB), and neoantigen load (NAL). In immune checkpoint inhibitor (ICI) treatment cohorts, low IRLS scores were associated with improved response rates to atezolizumab. Our findings suggest that the IRLS model serves as a novel biomarker for prognostic stratification and personalized therapeutic decision-making in breast cancer.