结构变异检测面临着两个方面的挑战：1）线性参考基因组难以清晰明确地表示基因组间结构差异的诸多形式；2）重复序列、测序错误以及比对错误严重影响测序序列在基因组上的准确定位。第三代测序技术凭借其超长读长的优势能够显著提高测序序列在基因组上定位的精度，并且基于图结构的参考基因组能够完整地表示基因组间的所有变异类型，从而提高检测的准确性和敏感性。利用第三代测序技术和组装图的优势，申请者前期参与了SMARTdenovo和wtdbg2的研发过程，主导了LRScaf的开发工作。基于以上开发中的经验积累，本项目将改进团队提出对第三代测序数据具有高容错性的模糊布鲁因图，增加红黑图的特性，研发基于第三代测序技术的结构变异检测算法，并在变异类型推断中引入最大似然估计，量化鉴定标准，降低检测的假阳性。本项目研发的算法将为基于高噪音长读长的结构变异检测提供新的方法学，预期在准确性和召回率上均显著优于现有同类算法。

Structural variations (SVs) detection faces two challenges: 1) linear reference genomes are difficult to clearly and unequivocally represent the many possible forms of structural differences among genomes; 2) repeated sequences, sequencing errors, and alignment errors seriously affect the accurate localization of sequencing sequences on the genome. Third-generation sequencing (TGS) technology can significantly improve the accuracy of sequencing sequence localization on the genome by virtue of its ultra-long read length, and reference genome based on the graph structure can represent all variant types among genomes in a complete manner, thereby improving the accuracy and sensitivity of SVs detection. Taking advantage of TGS technology and assembly maps, the applicant participated in the development process of SMARTdenovo and wtdbg2 in the early stage and led the development of LRScaf. Based on the experience accumulated in the above development, this project will improve the fuzzy-Bruijn Graph (FBG) which was introduced by our team with high errors tolerance for TGS data, add the characteristics of red and black graph, develop a SVs detection algorithm based on TGS technology, and introduce maximum likelihood estimation in variation type inference to quantify identification criteria and reduce the detection of false positives. The algorithm developed in this project will provide a new methodology for long noise reads on SVs detection and is expected to be significantly better than existing similar algorithms in accuracy and recall.