Abstract
Structural variants (SVs) play a critical role in shaping the diversity of the human genome, and their detection holds significant potential for advancing precision medicine. Despite notable progress in single-molecule long-read sequencing technologies, accurately identifying SV breakpoints and resolving their sequence remains a major challenge. Current alignment-based tools often struggle with precise breakpoint detection and sequence characterization, whereas whole-genome assembly-based methods are computationally demanding and less practical for targeted analyses. Neither approach is ideally suited for scenarios where regions of interest are predefined and require precise SV characterization. To address this gap, we introduce FocalSV, a targeted SV detection framework that integrates both assembly- and alignment-based signals. By combining the precision of local assemblies with the efficiency of region-specific analysis, FocalSV enables more accurate SV detection. FocalSV supports user-defined target regions and can automatically identify and expand regions with potential structural variants to enable more comprehensive detection. FocalSV is evaluated on 10 germline data sets and two paired normal-tumor cancer data sets, demonstrating superior performance in both precision and efficiency.