Actionable genetic landscape and dynamics in circulating DNA from 12,827 patients with metastatic breast cancer

Plasma cell-free DNA (cfDNA) is a sensitive, noninvasive tool for monitoring the genomic landscape, heterogeneity, and evolution of cancer, aggregating genetic information from multiple metastatic sites. Here, we study 15,564 cfDNA samples from 12,827 patients with metastatic breast cancer (MBC). We classify samples, deep-sequenced with Guardant360, into MBC subtypes using our trained random forest model. Most patients have variants associated with therapeutic resistance and targetability of approved and experimental therapies, including in PIK3CA, ESR1, PTEN, and AKT1. Frequencies of multiple actionable alterations differed from primary disease, and multiple variants per gene suggest independent resistant clones. The large cohort size enables uncovering clonal co-behaviors, including co-existence and exclusivity of alleles and correlation of allele fractions. cfDNA testing is now recommended in multiple advanced malignancies, including eligibility assessment for clinical trials. We envision this large collection of MBC cfDNA data as an invaluable resource for the design and planning of future clinical trials.