Abstract Double-strand break repair leaves recognizable footprints in the genome. Among the most specific are short sequences inserted at structural-variant (SV) junctions—templated insertions often attributed to polymerase-θ-mediated end joining (TMEJ). Yet common readouts based on exact string matches overlook sequence background, distance from the break, and the topological context of candidate templates, inflating false positives and blurring mechanistic interpretation. We present a scalable statistical framework that infers templated insertions with controlled specificity by modeling alignment-score distributions against a distance-adjusted, genome-wide empirical null. The approach explicitly accommodates imperfect copying and partitions candidate templates into four breakpoint-proximal configurations, capturing positional and strand relationships that are informative of mechanisms. Applied to large somatic and germline whole-genome cohorts, the method reveals that template usage spans all configurations but differs systematically across cellular contexts. A notable fraction of events reflect imperfect copying, consistent with error-prone synthesis, whereas one configuration shows comparatively higher apparent fidelity—suggesting distinct biochemical routes within a broader TMEJ-like landscape. Configuration calls also stratify SV architecture: some are enriched in simple rearrangements while others localize to clustered, complex regions, indicating that local topology and repair pathway choice are linked. Beyond structure, configuration-specific burdens align with DNA-repair states and selected genotypes: contexts consistent with homologous-recombination deficiency show enrichment in particular configurations, while others display the opposite directionality, underscoring that “templated insertion” is not a single phenomenon but a family of related processes with diverging determinants. To enable cohort-scale analysis, we optimized the core alignment to produce full score matrices in a single pass and packaged the workflow into a containerized pipeline, yielding order-of-magnitude speedups and portable reproducibility. Together, these results establish a configuration-aware, statistically principled readout of templated insertions that is robust to sequence confounders and informative about mechanisms. Practically, the framework provides (i) a sharper lens for studying double-strand break repair in human samples, (ii) leads for repair-state biomarkers, and (iii) hypotheses connecting SV topology to polymerase usage. In doing so, it aims to move the field from anecdotal sequence sketches toward reproducible, cohort-scale inferences about double-strand break repair involving junctional insertions. Citation Format: Youyun Zheng, Gregory Raskind, Sophie Webster, Narmen Azazmeh, Haruna Tomono, Andrew Cherniack, David Lehotzky, Ron Solan, Antonia Kowalewski, Xavi Loinaz, Hansol Park, Vasuki N. Swamy, David Heiman, Samantha Van Seters, Saveliy Belkin, Sam Wiseman, Chunyang Bao, Luis A. Corchete Sanchez, Zachary Everton, Ryul Kim, Beomki Lee, Won-Chul Lee, Chip Stewart, Gengchao Wang, Brian P. Danysh, Young Seok Ju, Esther Rheinbay, Gad Getz, Rameen Beroukhim. Origins of structural variant junctional insertions across 8,000 TCGA whole genomes 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 3247.
Zheng et al. (Fri,) studied this question.