Abstract 2406: LossFinder: a novel algorithm for detecting PTEN homozygous deletions in cancer

Abstract Introduction: The PTEN gene, located at 10q23.31, is a tumor suppressor that regulates the PI3K/AKT signaling pathway. PTEN deletions and/or loss of protein are common in cancers, with The Cancer Genome Atlas (TCGA) reporting a prevalence of 10-85%, about half of which are due to copy number loss. PTEN homozygous deletions, more significant than hemizygous deletions, are key markers in cancer treatment, especially with the approval of the AKT inhibitor capivasertib. Homozygous PTEN deletions occurs in 4.28% of cancers, 5.08% in breast cancer, and 17.42% in prostate cancer. Detecting exon-level deletions in the PTEN gene is particularly challenging, given the short coding regions, the influence of tumor heterogeneity, and the effects of sample quality on Next-Generation Sequencing (NGS) data. To overcome these challenges, we developed LossFinder, a novel algorithm for detecting homozygous PTEN deletions. Methods: LossFinder is designed to detect homozygous gene copy number deletions in tumor NGS sequencing. The algorithm: 1) employs multi-depth correction and noise reduction to enhance the signal-to-noise ratio, mitigating the effects of tumor heterogeneity and experimental noise; 2) uses the circular binary segmentation (CBS) algorithm for precise breakpoint localization, improving detection of small exon-level copy number alterations; 3) assesses tumor purity and ploidy through mutation frequency and copy number alterations, restoring absolute copy numbers to distinguish homozygous from hemizygous deletions. We evaluated limits of detection (LoD) and precision using mixed tumor cell lines from ATCC. Additionally, clinical tumor samples were analyzed for PTEN deletion prevalence, with IHC used as the gold standard for comparison. Results: In cell line simulations, the algorithm demonstrated 100% specificity and 96.21% sensitivity when tumor purity was ≥30%. The LoD was achieved with 30% tumor purity, and precision with 35% tumor purity was 100%. In a cohort of 5,944 clinical FFPE tissue samples from over 20 cancer types, the prevalence of PTEN homozygous deletions was 4.32% overall, 6.56% (12/183) in breast cancer, 6.52% (3/46) in prostate cancer, and 3.77% (116/3078) in lung cancer, matching TCGA data for breast cancer but with a lower prevalence in prostate cancer, likely due to sample size limitations. Whole PTEN gene homozygous deletions accounted for 40% of PTEN homozygous deletion-positive samples. IHC validation in 97 clinical samples showed 85.71% consistency with NGS. Conclusions: The LossFinder algorithm offers high sensitivity and specificity in detecting PTEN homozygous deletions. It accurately identifies both whole and partial homozygous deletions, showing a high level of consistency with IHC. Citation Format: Ting Kuang, Chenglin Liu, Zhihong Zhang, Zhou Zhang, Fujun Qiu, Yuzi Zhang. LossFinder: a novel algorithm for detecting PTEN homozygous deletions in cancer [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 2406.