Integration of Radiomics and Immune-Related Genes Signatures for Predicting Axillary Lymph Node Metastasis in Breast Cancer

Abstract

BackgroundTo develop a radiogenomics nomogram for predicting axillary lymph node (ALN) metastasis in breast cancer and reveal underlying associations between radiomics features and biological pathways. Materials and methodsThis study included 1062 breast cancer patients, 90 patients with both DCE-MRI and gene expression data. The optimal immune-related genes and radiomics features associated with ALN metastasis were firstly calculated, and corresponding feature signatures were constructed to further validate their performances in predicting ALN metastasis. The radiogenomics nomogram for predicting the risk of ALN metastasis was established by integrating radiomics signature, immune-related genes (IRG) signature, and critical clinicopathological factors. Gene modules associated with key radiomics features were identified by weighted gene co-expression network analysis (WGCNA) and submitted to functional enrichment analysis. Gene set variation analysis (GSVA) and correlation analysis were performed to investigate the associations between radiomics features and biological pathways. ResultsThe radiogenomics nomogram showed promising predictive power for predicting ALN metastasis, with AUCs of 0.973 and 0.928 in the training and testing groups, respectively. WGCNA and functional enrichment analysis revealed that gene modules associated with key radiomics features were mainly enriched in breast cancer metastasis-related pathways, such as focal adhesion, ECM-receptor interaction, and cell adhesion molecules. GSVA also identified pathway activities associated with radiomics features such as glycogen synthesis, integration of energy metabolism. ConclusionThe radiogenomics nomogram can serve as an effective tool to predict the risk of ALN metastasis. This study provides further evidence that radiomics phenotypes may be driven by biological pathways related to breast cancer metastasis.