结构变异与疾病特别是癌症密切相关，因此结构变异检测在疾病诊断和药物设计等方面都具有重大的应用价值，也是生物大数据分析的重要任务之一。而不同测序技术产生的长读数和短读数各有特点，本项目将通过挖掘分析长短读数的互补优势，设计高效准确的结构变异检测方法。本项目拟从四个方面开展研究：（1）研究设计以检测结构变异为优化目标的长读数和基因组参考序列之间的比对算法；（2）研究如何利用已知局部区域特征，设计基于长短读数的局部组装方法以检测结构变异；（3）研究设计在深度学习框架下利用长短读数对结构变异进行检测的方法，特别是比对结果转化为适合深度学习模型输入层的方法；（4）研究现有结构变异检测方法的融合策略，进一步提高检测结果的准确性。本项目的实施将为生物学和医学等研究提供重要帮助。

Structural variation is closely related to diseases, especially cancer. Therefore, structural variation detection has great application value in disease diagnosis and drug design. It is also one of the important tasks of biological data analysis. The long and short reads produced by different sequencing technologies have their own characteristics. This project will design efficient and accurate methods to detect structural variation by mining and analyzing the complementary advantages of long and short reads. This project aims to carry out research from the following four aspects: (1) The project will develop new alignment algorithm between long reads and genome reference, which aims at identifying structural variation; (2) By analyzing local region’s characteristics, the project will develop the local assembly method to identify the location and type of structural variation; (3) This project will design method of detecting structural variation using long and short reading under the framework of deep learning, especially how to transfer the alignments into the input layer of deep learning model. (4) For further improving the accuracy of detection results, the fusion strategy of the detection results from the existing structure variation detection methods will be studied. The implementation of this project will provide important help for biology and medical researches.