Prognostic Biomarkers in Breast Cancer via Multi-Omics Clustering Analysis

Breast cancer (BC) is a highly heterogeneous disease with diverse molecular subtypes, which complicates prognosis and treatment. In this study, we performed a multi-omics clustering analysis using the Cancer Integration via MultIkernel LeaRning (CIMLR) method on a large BC dataset from The Cancer Genome Atlas (TCGA) to identify key prognostic biomarkers. We identified three genes-<i>LMO1</i>, <i>PRAME</i>, and <i>RSPO2</i>-that were significantly associated with poor prognosis in both the TCGA dataset and an additional dataset comprising 146 metastatic BC patients. Patients' stratification based on the expression of these three genes revealed distinct subtypes with markedly different overall survival (OS) outcomes. Further validation using almost 2000 BC patients' data from the METABRIC dataset and RNA sequencing data from therapy-resistant cell lines confirmed the upregulation of <i>LMO1</i> and <i>PRAME</i>, respectively, in patients with worse prognosis and in resistant cells, also suggesting their potential role in drug resistance. Our findings highlight <i>LMO1</i> and <i>PRAME</i> as potential biomarkers for identifying high-risk BC patients and informing targeted treatment strategies. This study provides valuable insights into the multi-omics landscape of BC and underscores the importance of personalized therapeutic approaches based on molecular profiles.