Evaluating Multi-trait PRS Methods on Cancer Phenotypes in PLCO

Principal Investigator

Name
Phillip Kraft

Degrees
Ph.D.

Institution
National Cancer Institute

Position Title
Senior Investigator

Email
phillip.kraft@nih.gov

About this CDAS Project

Study
PLCO (Learn more about this study)

Project ID
PLCO-2004

Initial CDAS Request Approval
Dec 16, 2025

Title
Evaluating Multi-trait PRS Methods on Cancer Phenotypes in PLCO

Summary
The development of Polygenic Risk Scores (PRS) derived from large-scale biobank analyses holds immense promise for improving disease risk stratification. However, the clinical utility of newly derived and state-of-the-art PRS models, particularly those leveraging multi-trait approaches such as PRSmix, must be rigorously validated in independent, diverse cohorts. The Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial (PLCO) cohort, with its extensive longitudinal follow-up and diverse phenotype data, presents a critical, well-characterized validation dataset. We will apply state-of-the-art PRSs, generated from multi-ancestry data in large-scale biobank studies in the NIH-funded PRIMED consortium, to the genotyped PLCO participants. For all outcomes, including traits and diseases, such as BMI, T2D, CAD, common cancers, we will calculate the PRS-specific prediction score for each individual. Performance will be quantified using the Area Under the Receiver Operating Characteristic curve (AUC) for binary outcomes (e.g. T2D, CAD, Cancers) and the coefficient of determination (R2) for continuous outcomes (BMI), and for the proportion of variance explained for binary outcomes (using an appropriate pseudo-R2). Our goal is to determine the utility of the novel PRS(s) derived via PRSmix and other methods in PLCO and characterize their performance.

Aims

Primary Aim: Quantify the Predictive Performance of Novel PRSs for Continuous and Binary Outcomes.

• Apply the pre-existing, optimized Polygenic Risk Scores (PRSs) for diseases and traits, such as BMI, T2D, CAD, Breast Cancer, and Prostate Cancer, to the genotyped individuals in the PLCO cohort.
• For the continuous traits (e.g., BMI), calculate the proportion of phenotypic variance explained by the PRS using the coefficient of determination (R2).
• For the binary traits (e.g., T2D, CAD, Breast Cancer, and Prostate Cancer), determine the discriminatory accuracy of the PRS-only model using the Area Under the Receiver Operating Characteristic curve (AUC) and quantify the proportion of liability variance explained using an appropriate pseudo- R2 measure.

Collaborators

Phillip Kraft National Cancer Institute
Jayati Sharma National Cancer Institute
Sonja Berndt National Cancer Institute