Multi-Machine Learning Elucidates Clinical Potential of Epithelial-Mesenchymal Transition-Associated Long Non-Coding RNAs in Breast Cancer Progression.

Breast carcinoma (BRCA) involves multiple molecular markers, including epithelial-mesenchymal transition (EMT), which induce cell migration. However, the specific impact of long non-coding RNAs (lncRNAs) on EMT in BRCA remains uncertain. In this study, a prognostic model was constructed using EMT-associated lncRNAs (EALs), with utilization of integrative machine learning algorithms. The optimal model consisted of 15 EALs, with an AUC of 0.89 at 5 years, showing its potential as a plausible biomarker for BRCA. Among high-risk individuals, a significant increase in pathways linked to the preservation of equilibrium and immune defense was observed. Moreover, it was indicated that immunotherapy elicited negative responses in this group. Somatic mutations displayed higher TP53 rates in high-risk patients and increased CDH1/PIK3CA in low-risk ones. Notably, AC055854.1 and MIR205HG, important EALs in the model, probably regulate BRCA development through the lncRNA-microRNA-mRNA axis. Spatial transcriptome analysis revealed higher expression levels of EALs and high-risk related genes in ductal carcinoma in situ (DCIS), invasive mixed ductal/lobular carcinoma (IDC), and triple-negative BRCA (TNBC) than in breast metastasis (BMS) samples. And neutrophils were exclusively observed within the tumor microenvironment (TME) of BMS. All these findings emphasized EALs' value in revolutionizing clinical decision-making for personalized treatment strategies in BRCA cases.