Abstract
<title>Abstract</title> <p> Circulating tumor DNA (ctDNA) mutation profiling provides a minimally invasive strategy for identifying prognostic biomarkers in breast cancer. Whole-genome sequencing (WGS) data from 349 patients with breast cancer were analyzed to characterize plasma ctDNA mutation profiles, with 1,124 healthy donors used for background variant filtering. ctDNA profiling identified 10,380 plasma-derived variants involving 5,930 genes across the breast cancer cohort, comprising 8 mutation classes, predominantly represented by missense mutations, frameshift deletions, and nonsense mutations. Recurrently mutated genes included <italic>MUC12</italic> , <italic>HRNR</italic> , <italic>OBSCN</italic> , <italic>TTN</italic> , <italic>EPPK1</italic> , <italic>HMCN2</italic> , <italic>RYR1</italic> , <italic>LRP1</italic> , and <italic>MUC16</italic> . Subtype-associated mutation analysis identified representative subtype-enriched mutated genes, including <italic>TTN</italic> and <italic>OBSCN</italic> in luminal A, <italic>MUC12</italic> and <italic>AMZ1</italic> in luminal B, <italic>ARHGEF17</italic> and <italic>DNHD1</italic> in HER2-enriched disease, and <italic>LARGE2</italic> and <italic>COL7A1</italic> in triple-negative breast cancer.Tissue-level mutation profiles and transcriptomic survival cohorts were used to prioritize and validate ctDNA-derived prognostic biomarkers. A prognostic model integrating <italic>NFKBIA</italic> , <italic>PDCD1</italic> , <italic>ARID1B</italic> , <italic>CFB</italic> , <italic>JAK1</italic> , <italic>MUC4</italic> , and <italic>FCGBP</italic> stratified breast cancer patients into distinct survival-risk groups and was strongly associated with overall survival (HR = 2.31, 95% CI: 1.64–3.26, <italic>P</italic> < 0.0001). Mutation-based classification of these seven genes further revealed distinct survival patterns across genetically defined subgroups ( <italic>P</italic> = 0.011). In conclusion, ctDNA mutation profiling demonstrates the utility of plasma-derived genomic features and a seven-gene prognostic score in risk stratification of 1,385 breast cancer patients within 7 datasets, which may support future liquid biopsy-based biomarker discovery and prognostic assessment. </p>