PROJECT 3: Summary The ongoing WISDOM risk-based breast cancer screening and prevention trial, which started in 2016, is a ground-breaking comparative effectiveness study to establish whether personalized breast cancer screening based on individual risk is non-inferior to annual screening in detecting advanced cancers, while reducing the potential harms of screening. However, the current WISDOM risk model has only moderate predictive power, which therefore offers high potential for improvement. Since the advent of the WISDOM trial, there have been substantial innovations in risk prediction: 1) deep-learning mammographic artificial intelligence (AI) methods can extract numerous mammographic features beyond breast density; 2) ongoing genome-wide association studies (GWAS) have improved the understanding of genetic susceptibility for breast cancer subtypes and for populations of non-European descent; and 3) molecularly defined subtypes of fast- and slow-growing cancers are being used as more clinically relevant endpoints than the presence or absence of any breast cancer. The overall goal of Project 3 is to develop and improve integrated risk models combining advances in deep- learning mammographic AI, germline genetics, and clinical factors for predicting molecular subtypes of breast cancer. This next-generation predictive model will be incorporated into the ongoing WISDOM study platform for risk-based screening and prevention. We will achieve this through 3 Specific Aims: We will 1) Develop a deep-learning mammographic AI model to predict risk for molecular subtypes of fast- and slow-growing breast cancer in large, diverse cohorts; 2) Develop and validate the next-generation WISDOM risk model by combining the Aim 1 mammographic AI model and Project 2 polygenic risk score (PRS) model with the latest clinical risk models to predict risk for molecular subtypes of fast- and slow-growing breast cancer; and 3) Transfer this mammographic AI model from standard 2D mammography to newer 3D tomosynthesis in a diverse cohort. The Project 3 team has extensive expertise with developing integrated risk models. Furthermore, the research will benefit from access to 5 large, diverse study cohorts with >200,000 women with >14,000 breast cancer cases in which these risk features will be available. The final model will be used to update CISNET simulation model in Project 4 that will comprehensively estimate benefits, harms, costs, and health value outcomes from the developed models. Guided by the updated WISDOM simulation and multidisciplinary group consensus, Projects 3 and 4 will identify a unified risk-based approach to be implemented in the WISDOM study as the next-generation WISDOM risk platform. These new models and simulations are expected to substantially improve the accuracy and clinical utility of the current WISDOM risk model, enabling greater reductions in breast-cancer incidence, related morbidity and mortality, and harms related to over-screening.