肽段序列鉴定是蛋白质组学完整性分析、基因组注释完善与校正等生物信息学研究的基础性工作。由于质谱数据的爆炸性增长，肽段测序目前面临计算时间长、测序规模小等瓶颈，迫切需要高性能计算技术的支持。本项目拟研究大规模的肽段测序及其基于CPU+GPU+FPGA混合异构系统的并行处理关键技术。重点研究以下问题：1）针对现有数据库搜索算法可扩展性受限问题，拟提出基于中心关联查找的序列数据库并行搜索算法，以及采用内存预分配和复用策略设计并行肽段-质谱匹配算法。2）针对肽段从头测序方法计算时限长问题，拟设计基于图论模型的谱峰连接图的并行构建算法，在此基础上，提出细粒度的候选肽段并行评分算法。3）从存储方案、资源管理和任务调度等方面，研究混合异构系统上肽段测序应用的并行性能优化问题。本项研究将为蛋白质组学提供高效的肽段实时测序工具，也为设计其他数据密集型应用的混合异构系统解决方案提供新的思路。

Peptide sequence identification is a fundamental research in bioinformatics such as proteomics integrity analysis, genome annotation improvement and correction. Due to the explosive growth of the mass spectrometry data scale, peptide sequencing methods suffer from bottlenecks such as long calculation time and small sequencing scale, which urgently need the support of high-performance computing technology. This project investigates large-scale peptide sequencing and its parallel processing key technologies based on CPU + GPU + FPGA hybrid heterogeneous system. We will mainly study the following three types of key problems, 1) Aiming at the limited scalability of existing database search algorithms, a parallel search algorithm for sequence database based on central association search is proposed, and a parallel peptide-mass spectrometry matching algorithm is designed using memory pre-allocation and multiplexing strategy. 2) Aiming at the problem of calculating the time limit of the peptide de novo sequencing method, a parallel construction algorithm of spectrum peak connection graph based on graph theory model is proposed. Based on the above, a fine-grained parallel scoring algorithm for candidate peptides is proposed. 3) From the aspects of storage solutions, resource management and task scheduling, the parallel performance optimization of peptide sequencing applications on hybrid heterogeneous systems is studied. This project will provide an efficient real-time tool for peptide sequencing. It will also provide new ideas for other data-intensive applications on hybrid heterogeneous system.