Identification of VWA5A as a novel biomarker for inhibiting metastasis in breast cancer by machine-learning based protein prioritization

Distant metastasis is the leading cause of death in breast cancer (BC). The timing of distant metastasis differs according to subtypes of BCs and there is a need for identification of biomarkers for the prediction of early and late metastasis. To identify biomarker candidates whose abundance level can discriminate metastasis types, we performed a high-throughput proteomics assay using tissue samples from BCs with no metastasis, late metastasis, and early metastasis, processed data with machine learning-based feature selection, and found that low VWA5A could be responsible for shorter duration of metastasis-free interval. Low expression of VWA5A gene in METABRIC cohort was associated with poor survival in BCs, especially in hormone receptor (HR)-positive BCs. In-vitro experiments confirmed tumor suppressive effect of VWA5A on BCs in HR+ and triple-negative BC cell lines. We found that expression of VWA5A can be assessed by immunohistochemistry (IHC) on archival tissue samples. Decreasing nuclear expression of VWA5A was significantly associated with advanced T stage and lymphatic invasion in consecutive BCs of all subtypes. We discovered lower expression of VWA5A as the potential biomarker for metastasis-prone BCs, and our results support the clinical utility of VWA5A IHC, as an adjunctive tools for prognostication of BCs.