Abstract 1910: Allelic imbalance for early detection of breast cancer in high depth whole genome sequencing

Abstract Liquid biopsy for circulating tumor DNA (ctDNA) offers potential for noninvasive early cancer detection. However, existing mutation- and methylation-based multi-cancer early detection tests have poor sensitivity in low-shedding, high-aneuploidy tumor types like breast cancer. We previously developed an ultrasensitive tumor-informed liquid biopsy approach that used B-allele frequency (BAF) to detect cancer aneuploidy. Our BAF classifier, which measures allelic imbalance from the millions of single nucleotide polymorphisms (SNPs) across the genome, demonstrated high sensitivity for minimal residual disease in triple-negative breast cancer (TNBC) using only standard whole genome sequencing (WGS). To expand our BAF approach to the tumor-naïve early detection context, we sought to enhance aneuploidy signal through (i) high-depth plasma WGS (150-200x) using Ultima Genomics to better interrogate allelic imbalance at individual SNP sites and (ii) long-read sequencing of normal tissue to phase A and B haplotypes, allowing for the aggregation of signal from large regions of neighboring SNPs. We characterized two metrics of allelic imbalance: BAF and a fragment length score (FRAG) characterized by cancer-specific differences in fragment lengths between phased alleles. To assess the performance of integrated BAF and FRAG signals for ctDNA detection, we collected a pilot cohort of BRCA-mutant, early-stage breast cancer patients with 7 pre-treatment and 28 post-surgery plasma samples from patients with no documented recurrence. ctDNA was overtly detected in pre-treatment samples from a stage III HR+/HER2- breast cancer and a stage II TNBC. We further sought to evaluate performance in stage I breast cancer, a particularly challenging setting for ctDNA detection through existing early detection approaches. We detected a stage I pT1c HR+/HER2- breast cancer and a stage I pT1b (<1 cm) TNBC at 100% specificity against post-surgery samples, demonstrating the potential of our classifier for non-invasive early cancer detection. Overall area under the curve (AUC) between pre-treatment (n=7) and post-surgery (n=28) samples was 0.86 in this challenging cohort that included 3 HR+/HER2- stage I (2 pT1a, 1pT1b) breast cancer samples. Longitudinal analysis of serial plasma samples throughout neoadjuvant treatment and following surgery demonstrated the ability of our tumor-naïve classifier to track therapeutic response. Our method demonstrates promising potential for early cancer detection and tumor-naïve monitoring of treatment response in breast cancer patients, including those with very small tumors. We expect this method to extend beyond breast cancer to the many highly aneuploid cancer types where ctDNA detection through conventional methods has been limited. Additional samples will further validate our approach. Citation Format: Laura Andersen, Konner Nelson, Julia Ah-Reum An, Pedram Razavi, Mark Robson, Dan Landau, Nicolai Birkbak, Adam Widman. Allelic imbalance for early detection of breast cancer in high depth whole genome sequencing [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 1910.