Evaluation\nof a Proteomics-Guided Protein Signature\nfor Breast Cancer Detection in Breast Tissue

The distinction between noncancerous and cancerous breast\ntissues\nis challenging in clinical settings, and discovering new proteomics-based\nbiomarkers remains underexplored. Through a pilot proteomic study\n(discovery cohort), we aimed to identify a protein signature indicative\nof breast cancer for subsequent validation using six published proteomics/transcriptomics\ndata sets (validation cohorts). Sequential window acquisition of all\ntheoretical (SWATH)-based mass spectrometry revealed 370 differentially\nabundant proteins between noncancerous tissue and breast cancer. Protein–protein\ninteraction-based networks and enrichment analyses revealed dysregulation\nin pathways associated with extracellular matrix organization, platelet\ndegranulation, the innate immune system, and RNA metabolism in breast\ncancer. Through multivariate unsupervised analysis, we identified\na four-protein signature (OGN, LUM, DCN, and COL14A1) capable of distinguishing\nbreast cancer. This dysregulation pattern was consistently verified\nacross diverse proteomics and transcriptomics data sets. Dysregulation\nmagnitude was notably higher in poor-prognosis breast cancer subtypes\nlike Basal-Like and HER2 compared to Luminal A. Diagnostic evaluation\n(receiver operating characteristic (ROC) curves) of the signature\nin distinguishing breast cancer from noncancerous tissue revealed\narea under the curve (AUC) ranging from 0.87 to 0.9 with predictive\naccuracy of 80% to 82%. Upon stratifying, to solely include the Basal-Like/Triple-Negative\nsubtype, the ROC AUC increased to 0.922–0.959 with predictive\naccuracy of 84.2%–89%. These findings suggest a potential role\nfor the identified signature in distinguishing cancerous from noncancerous\nbreast tissue, offering insights into enhancing diagnostic accuracy.