Abstract 6542: Detection of androgen receptor splice variants from clinical sequencing

Abstract Androgen receptor (AR) splice variants (AR-Vs) have been widely studied on its important role in prostate cancer (PC) progression. In particular, AR-Vs lacking ligand binding domains, such as AR splice variant 7 (AR-v7) that links exon 3 and cryptic exon 3 or AR variant that losses exons 5 to 7 (ARv567es), have been demonstrated as one of the resistant mechanisms to androgen deprivation therapy in PC. Despite its clinical importance, the current algorithms to detect fusions such as TopHat-Fusion have limitation to detect intra-chromosomal rearrangements including AR-Vs. In this respect, we propose a novel approach to identify AR-Vs from targeted RNA sequencing (RNA-seq) dataset. In short, our two-step approach first gathers soft-clipped or divided reads that are adjacent to the splicing sites of AR-Vs and then yields the number of splitting reads that support the existence of AR-Vs. Next, the algorithm selects the paired-end reads whose one side and the other side are mapped to the preceding and following exons, respectively. The final number of spanning reads are calculated following to the removal of low-quality reads. We validated the proposed approach using two large-scale, independent RNA-seq datasets of PC samples: 34 samples from Seoul National University Hospital (SNUH) and 558 samples from The Cancer Genome Atlas (TCGA) dataset. Overall, our approach successfully identified samples with putative AR-Vs, including AR-v7 and ARv567es. In addition, our approach successfully detected AR-v7 with reliable number of supporting reads, from the RNA-seq dataset of AR-v7-expressing cell line. Finally, statistical analyses of 34 SNUH PC patients suggested potential detection threshold in clinical sequencing settings (at least 10 split reads and 1% proportion). In the SNUH dataset, The AR-v7 positive group, which included 6 patients (17.1%) with values above the threshold, had higher AR-v7 expression levels than the negative group (p=4.8E-04). The similar pattern was also found in the analysis of TCGA dataset. Owing to its flexibility, we also confirmed that the developed approach can be used to detection of other AR-Vs, including AR-v3 and ARv567es. We expect that the further in-depth analyses including larger samples and clinical outcomes can discover clinical applicability of AR-Vs. Citation Format: Suhyun Hwangbo, Sungyoung Lee, Sheehyun Kim, Hongseok Yun. Detection of androgen receptor splice variants from clinical sequencing [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 6542.