Fig. 1

Schematic of the study design and analytical methods. a A nested case–control study was constructed within the EPIC-Heidelberg cohort. Blood samples and lifestyle information were collected from apparently healthy participants upon enrolment. Participants who were diagnosed with breast cancer over the course of follow-up were matched on a one-to-one basis with individuals who were observed to be cancer-free over the study period. Buffy coat lysates derived from blood samples collected at enrolment were analysed by reduced representation bisulphite sequencing (RRBS). b Fivefold nested cross-validation was used to train and evaluate classifier models for their ability to discriminate individuals who developed breast cancer from those who were cancer-free over the follow-up period. The best-performing model was selected and trained on the full cross-validation dataset of 340 pairs, to finalize model parameters. The final model was used to predict case–control status in a held-out validation set of 68 matched pairs. Differential methylation analyses and functional enrichment analyses were conducted in parallel. This figure uses images from Servier Medical Art licensed under a Creative Commons Attribution 3.0 Unported License (https://creativecommons.org/licenses/by/3.0/)