Fig. 1: Overview of experimental design and datasets.

We included women with breast cancer who had at least one poor prognostic feature and recruited both hospital-based controls and controls from the general population. The discovery set was further subdivided into a training set and internal validation set and was used to derive the WID-BC-index. For a subset of individuals in the internal validation set, matched buccal samples and SNP data were available. The WID-BC-index was evaluated in the external validation set. Additional datasets for evaluation of the WID-BC-index consisted of cervical samples from individuals with endometrial or ovarian cancer, or BRCA mutation carriers, as well as breast tissue samples from healthy controls, BRCA mutation carriers, or normal tissue adjacent to triple-negative breast cancer. Predictive assessment of the WID-BC-index was carried out using samples from a Swedish Biobank which included cervical smears from healthy controls and women who were diagnosed with breast cancer 1 to 5 years after sample collection. Source data are provided as a Source Data file.