MOTIVATION: Structural variants (SVs) influence gene regulation, disease progression, and diagnostics, yet integrating SV calls across platforms remains difficult due to inconsistent annotations, limited merging flexibility, and fragmented workflows. Ambiguous breakend (BND) annotations, which comprise many variant calls, are often discarded or misclassified, hindering variant characterization. Existing tools lack advanced merging operations essential for precise identification of disease-specific or somatic variants across samples or patient groups. Additionally, current SV analysis pipelines require extensive manual intervention and complex parameter tuning, compromising reproducibility and scalability. Addressing these gaps is crucial for improving the accuracy, interpretability, and clinical utility of SV analyses. RESULTS: We developed OctopuSV and TentacleSV to address these long-standing challenges in SV analysis. OctopuSV features a specialized BND correction module that converts ambiguous BND annotations into canonical SV types, recovering important variants that are often overlooked by existing tools. Additionally, it provides advanced set operations (difference, complement, custom-defined) that enable sophisticated variant filtering without programming expertise, critical for identifying tumor-specific SVs or variants unique to specific sample groups. TentacleSV completes our solution by automating the entire SV analysis process from raw sequencing data to high-confidence callsets, ensuring consistency and reproducibility across projects. Benchmarking across short-read and long-read platforms showed superior F1 score, complete SV type consistency compared to existing tools. Our framework enables experimental biologists and clinical researchers to perform sophisticated analyses ranging from cancer subtype-specific SV identification to multi-sample comparative studies without requiring specialized programming skills. AVAILABILITY AND IMPLEMENTATION: All codes are available at https://github.com/ylab-hi/OctopuSV; https://github.com/ylab-hi/TentacleSV.

MOTIVATION: Neoantigens, tumor-specific protein fragments, are invaluable in cancer immunotherapy due to their ability to serve as targets for the immune system. Computational prediction of these neoantigens from sequencing data often requires multiple algorithms and sophisticated workflows, which are currently restricted to specific types of variants, such as single-nucleotide variants or insertions/deletions. Nevertheless, other sources of neoantigens are often overlooked. RESULTS: We introduce ScanNeo2 an improved and fully automated bioinformatics pipeline designed for high-throughput neoantigen prediction from raw sequencing data. Unlike its predecessor, ScanNeo2 integrates multiple sources of somatic variants, including canonical- and exitron-splicing, gene fusion events, and various somatic variants. Our benchmark results demonstrate that ScanNeo2 accurately identifies neoantigens, providing a comprehensive and more efficient solution for neoantigen prediction. AVAILABILITY AND IMPLEMENTATION: ScanNeo2 is freely available at https://github.com/ylab-hi/ScanNeo2/ and is accompanied by instruction and application data.

Semantic segmentation of aerial images plays an important role in urban area monitoring. But the diversity of buildings makes segmentation a hard task. To detect buildings from aerial images more precisely, this paper proposes a pixel-level segmentation method, named Attention Residual U-net (ARU-net). ARU-net adds two major part into the framework of U-net, i.e. attention path and residual connection, focusing on feature reuse. Attention path utilizes attention mechanism to capture spatial feature details. Residual connection implies the semantic information flow through a 1×1 convolution similar to the residual form. ARU-net can be trained end-to-end. Experiments are conducted to evaluate the effectiveness of the proposed model on the Inria Aerial Image Labeling Dataset. Results indicate that ARU-net outperforms other baselines with an accuracy of 93.84% and intersection over union (IoU) of 60.90%.