Abstract 3754: ecDNAInspector nominates high-confidence structural predictions of extrachromosomal DNA for downstream analysis

Abstract Extrachromosomal DNA (ecDNA) are circular, megabase-scale regions of the genome found outside canonical chromosomes and characterized by high copy-number and frequent oncogene amplification. They are implicated in tumorigenesis, therapeutic resistance, and shortened survival for patients across multiple cancer types. Existing tools to identify and predict ecDNA structures are limited to low throughput, high sensitivity enrichment-based assays or high throughput, low sensitivity computational methods using whole-genome sequencing data. Recent advances in long-read sequencing offer potential improvements in ecDNA detection but such data are currently limited. Evaluating the impact of ecDNA in relation to clinical characteristics requires large cohorts such as The Cancer Genome Atlas (TCGA) or the International Cancer Genome Consortium (ICGC). Hence, there is value in leveraging short-read sequencing data from these existing cohorts while improving confidence in computational predictions of ecDNA presence and structure. Here we present ecDNAInspector, a computational tool that enables systematic selection of high-confidence ecDNA predictions from sequencing data. Each ecDNA prediction is characterized using structure-based metrics: an extreme cycle size Boolean, a breakpoint mapping error Boolean, and adapted classification metrics that assess the frequency of paired ends between neighboring cycle segments with orthogonal structural variant call support. ecDNA predictions are clustered by these metrics and the user can select high-confidence groups. A subsequent filtering step enables rescue of additional ecDNA predictions based on user-defined thresholds. After selection of a final high-confidence group of ecDNA predictions, ecDNAInspector offers modules for structural comparison and analysis of ecDNA within a cohort, including inspection of structural conservation (i.e. maintenance of similar sequence) and feature comparisons across tumors. Applying ecDNAInspector to a breast cancer cohort, we nominated a subset of high-confidence ecDNA with strong orthogonal structural variant support and additional indicators of high quality. We identified both base pair and oncogene inclusion conservation within breast cancer subtype groups, and noted significant variations in structural metrics across subtype groups. Since structure often indicates origin and dictates function, these differences suggest potential biological differences in ecDNA genesis and evolution across breast cancer subtypes. Thus, ecDNAInspector provides a systematic method for selecting high-confidence ecDNA predictions for downstream structural analysis, enabling improved utilization of existing sequencing data in valuable clinical cohorts. Citation Format: Sophia J. Pribus, Kathleen E. Houlahan, Lise Mangiante, Christina Curtis. ecDNAInspector nominates high-confidence structural predictions of extrachromosomal DNA for downstream analysis [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 3754.