针对新一代测序平台产生的含有"indel（插入删除）"、"structural variation（结构变异）"错误的大量的read（序列）与大规模参考基因组比对的问题，研究建立多参数组合优化的序列比对（read alignment）模型，设计优化的序列比对串行算法；提出高效的序列比对并行计算方法、CPU/GPU结构多级存储数据分配方法和集群节点负载均衡的任务调度策略；在保持比对结果质量的前提下，设计将short read（短序列）快速定位到大规模参考基因组的粗粒度和细粒度混合并行的all-mappers比对算法；提出比对结果质量（精度、敏感度和召回率）更高、花费时间更少的将long read（长序列）定位到大规模参考基因组的粗粒度和细粒度混合并行的best-mapper和all-mappers比对算法。研究成果将促进序列比对并行算法的发展，为生物大数据分析应用研究提供高性能算法基础。

Aiming at the problem of aligning the large-scale reference genome and the large numbers of reads (sequences) with the "indel" and "structural variation" errors generated by the next generation sequencing platform, this project investigates to construct the optimized read alignment model with combined multiple parameters,design the optimized sequential algorithms for read aligment, propose the efficient parallel computing method of read alignment, the multi-level storage data allocation method, and load-balancing task scheduling strategy on the CPU/GPU architectures. Furthermore, this project investigates to design the all-mappers alignment algorithm with hybrid coarse-grained and fine-grained parallelism to locate the short reads to the large-scale reference genome on the premise of maintaining the quality of alignment results, propose the best-mapper and all-mappers alignment algorithms with hybrid coarse-grained and fine-grained parallelism to locate the long reads to the large-scale reference genome with higher quality (higher accuracy, sensitivity and recall) and less running time. The research achievements will promote the development of parallel algorithms for read alignments, and provide the basic high performance algorithms for the application research of biological data analysis.