2026 Sequencing Strategies for Population and Cancer Epidemiology Studies (SeqSPACE) : Submission #5

Abstract Summary Elevation of prostate-specific antigen (PSA) can be indicative of prostate cancer, but its use for population-based screening remains controversial. The same PSA level can have different interpretations depending on genetic predisposition, potentially leading to missed cases or overdiagnosis. Here we present an approach for obtaining genetically informed PSA values and validate it in the FinnGen biobank. Unlike previous methods that remove the component of PSA captured by a genome-wide polygenic score (PGS), we explore PGS partitioning to provide a more precise correction with fewer off-target genetic effects on prostate cancer risk. We first found that although the cis region explains only 20% as much variation in PSA as a genome-wide PGS, PSA adjusted by the cis-region PGS predicted prostate cancer nearly as effectively as with genome-wide PGS adjustment (OR=5.92, P=1.09e-444, AUC=0.840 vs. OR=6.07, P=3.52e-453, AUC=0.837). Both models improved upon observed pre-diagnostic PSA (OR=5.83, P=3.05e-440, AUC=0.838). Next, we partitioned the PGS into linkage disequilibrium blocks to identify trans loci that enhance the accuracy of genetic adjustment in the Prostate Cancer Prevention Trial (PCPT) and the Selenium and Vitamin E Cancer Prevention Trial (SELECT). Applying the resulting filtered PSA score to FinnGen yielded an even stronger association with prostate cancer than genome-wide adjustment, particularly in men with PSA ≤ 4 ng/ml (OR=6.07, P=5.42e-182, AUC=0.783 vs. OR=5.87, P=3.35e-179, AUC=0.779). This work demonstrates the importance of carefully selecting genetic signals when recalibrating noncausal disease biomarkers.