Abstract Background: Structural variants (SVs), including fusions, translocations, large insertions/deletions, inversions and duplications, are highly tumor-specific genomic events and powerful biomarkers for molecular residual disease (MRD) assessment. Detecting SVs with unknown partners using short-read sequencing remains challenging, especially at ultra-low variant fractions. We developed the Predicine DeepSEA SV detection algorithm to enable comprehensive discovery and sensitive tracking of multi-class SVs in tissue and plasma. Methods: Predicine DeepSEA integrates de novo assembly, refined BWA-based re-alignment, fragment-level molecular quantification, and combined split-read and paired-read evidence to detect and classify five SV subtypes with high accuracy. The workflow was applied to tumor tissue and matched longitudinal plasma samples from 30 patients with solid tumors or hematologic malignancies. PredicineBEACON MRD assays incorporated somatic SV breakpoints identified by whole-genome sequencing of high-tumor-fraction (20% TF) baseline samples. Personalized SV-based panels were then designed for targeted MRD tracking. Analytical sensitivity was evaluated using titration experiments in which patient-derived cfDNA was spiked into healthy donor cfDNA at defined dilutions down to 1 part per million (ppm). Results: DeepSEA identified diverse somatic SVs across all tumors, enabling construction of individualized MRD marker sets. In longitudinal plasma analysis, patient-specific SVs were quantifiable at extremely low variant fractions, and integration of multiple SV markers improved detection robustness compared with single-event tracking. In dilution studies, the PredicineBEACON assay achieved reliable MRD detection down to 1 ppm with high specificity, supported by fragment-level consensus filtering and multi-evidence breakpoint calls. SV-derived signals were detectable in both solid and hematologic cancers, underscoring the platform’s broad applicability. Conclusion: The Predicine DeepSEA pipeline provides sensitive and specific detection of structural variants in tissue and plasma. Incorporation of SVs into the PredicineBEACON MRD assay enables highly sensitive, personalized ctDNA tracking with detection limits approaching 1 ppm. This SV-based framework delivers a technically robust and broadly applicable platform for molecular monitoring across diverse cancer types. Citation Format: Kemin Zhou, Yong Huang, Abby He, Junmei Wang, Binggang Xiang, Xiaohong Wang, Pan Du.. Ultra-sensitive MRD detection through comprehensive structural variant profiling in solid and hematologic malignancies abstract. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 7825.
Zhou et al. (Fri,) studied this question.