Early-Onset Breast Cancer Risk: Age-Specific Genetic Susceptibility and Prediction in Diverse Populations

Study design:
International genome-wide association study.

Data Sources:

The project will leverage large-scale individual-level and summary-level data from the Confluence Project, an international collaboration integrating genetic and phenotypic data from more than 300,000 breast cancer cases and 300,000 controls across diverse ancestries (including of African, Asian, Hispanic/Latina, European and mixed ancestries). Confluence includes genome-wide genotype data (imputed to diverse reference panels, and including selected high- and moderate-risk mutations), harmonised epidemiological risk factors, and tumour characteristics.

Analyses:

Analyses will include development and validation of multi-ancestry PRS for overall and subtype-specific breast cancer, with a particular focus on early-onset disease. Although we expect that most genetic variants contributing to breast cancer risk are shared across ages, their effect sizes may differ by age at onset. To account for this, analyses will first construct PRS using SNPs identified from the overall multi-ancestry GWAS and then re-calibrate their effect sizes for younger-onset disease. Recalibration will include (a) re-estimating SNP weights using younger-onset cases, and (b) applying empirical-Bayes and hierarchical modelling approaches that borrow information across age groups while allowing age-specific deviations.
Both ancestry-specific and trans-ancestry PRS using approaches such as LDpred2 will be used, along with advanced Bayesian and machine-learning frameworks that integrate information across ancestries and age strata. Prospective cohort data from Confluence will serve as independent validation sets to assess predictive performance. These analyses will yield calibrated, ancestry-informed PRS across age groups and form the basis for evaluating joint effects of PRS and pathogenic variants in subsequent aims.
To assess interactions between PRS, carrier status in established predisposition genes, age, and family history, models that incorporate these factors simultaneously will be used. Final models will be selected using 10-fold cross-validation, maximising the area under the curve as a function of the L1 penalty. The same modelling strategy will be applied to develop ER-subtype–specific PRS.
Model estimates will be combined with country-specific age-specific incidence rates to calculate five-year and lifetime absolute risks (up to age 80) for carriers and non-carriers of pathogenic variants. Pointwise confidence intervals for absolute risks will be derived using parametric bootstrap samples.
To characterise the genetic architecture of early-onset (18–49 years) versus later-onset (50+ years) breast cancer, genetic correlations using LD score regression will be estimated. Analyses will be conducted for overall breast cancer and separately by ER, PR, and HER2 status to assess the extent to which subtype composition explains genetic differences by age at onset. Local genetic correlation analyses will be performed to identify genomic regions driving shared or distinct age-specific effects. Stratified LD score regression will be used to test whether shared versus age-specific loci are enriched for functional or regulatory annotations. Finally, differences in polygenicity and effect-size distributions between early- and later-onset disease will be evaluated using methods designed to estimate genome-wide effect-size architectures.

Impact:

By developing robust multi-ancestry PRS and quantifying how common variants modify risks associated with high- and moderate-risk genes, this research will enable more precise identification of younger women who could benefit from earlier screening or risk-reducing interventions. The work will also enhance understanding of the biological drivers of early-onset breast cancer by clarifying age-specific heritability and the genetic architecture of aggressive subtypes such as triple-negative disease. Ultimately, the project aims to enable risk prediction tools that are equitable, clinically relevant, and suitable for implementation in diverse populations. Outcomes will be directly translatable to risk-stratified screening strategies and personalised prevention efforts.

Research setting:

The PhD student will form part of the within the HER-CARE MSCA Doctoral Network, joining an international cohort of 15 PhD researchers focused on hereditary and early-onset breast cancer. The primary host institution (ICR) will provide training in genetic epidemiology. In addition, as part of the MSCA programme, the student will undertake:

•  a five-month secondment at John Hopkins University in the USA (hosted by Prof Nilanjan Chatterjee) that provide complementary experience in statistical genetics and data science;  and
• a three-month secondment at PHG Foundation in Cambridge, UK (hosted by Dr Laura Blackburn) that will provide complementary experience in clinical translation and policy evaluation.
  

In addition, the student will interact closely with investigators at the National Cancer Institute – the coordinating centre for Confluence – and other international investigators and collaborators. The research environment emphasises interdisciplinary training, open science, and international collaboration, ensuring that the student develops strong expertise in both methodological innovation and applied cancer research. The starting date will be before July 1, 2026.