Professor Kai Ye's team at Xi'an Jiaotong University has made advancements in the identification of germline and somatic structural variations, addressing the challenges of high false positives and difficulties in clinical implementation associated with traditional genome-wide comparative strategies. On March 22, their research findings were published in the journal Nature Biotechnology.
Studies on various genetic diseases and cancers require comparison of genomic variation differences among multiple samples. However, the commonly used stepwise strategy of "detect first, then compare" in the field requires multiple computational steps after genome detection. The complexity of multiple steps leads to rapid accumulation of errors and high false positives, making it difficult to accurately decipher germline and somatic structural variations.
Professor Ye's team proposed the SVision-pro algorithm, based on a "sequence-to-image" transformation strategy for multi-sample differential comparison. They broke through the traditional "detect first, then compare" strategy by transforming the detection and genotyping of structural variations from a sequence problem to an instance segmentation problem in image space. By directly comparing the visualized differences in sample sequencing, they achieved precise identification of germline and somatic structural variations with high accuracy and low false positives. This breakthrough provides critical technical support for identifying key pathogenic structural variations from large-scale disease-specific cohort data and clinical diagnostic data. It also offers a new perspective for the development of a bioinformatics framework based on "artificial intelligence+" in handling biological sequence big data.
Related Paper: https://doi.org/10.1038/s41587-024-02190-7
