中文摘要：

当前大部分重复体识别算法不是依靠于已经标识的重复体数据库就是定义重复体为两个最大长度的相似序列，而没有一个严格的定义来平衡重复体的长度和频率．针对这些问题文中提出了一种基于局部序列比对算法BLAST变型且支持空位的快速识别重复体的RepeatSearcher算法．算法通过定义重复体的精确边界运用逐步扩展调和序列来识别重复体．算法使用C．briggsae基因组序列作为测试对象，并与当前通用的重复体识别算法RECON以及新近的识别算法RepeatScout做了比较分析．结果表明RepeatSearcher使每一条重复体序列具有了精确的边界，而且相对其它算法在没有损失精度的情况下，缩短了算法的运行时间．

英文摘要：

Most existing methods of repeat identification either rely on annotated repeat databases or limit repeats to pairs of similar sequences that are maximal in length. And there is no an exact definition to correctly balances the importance of the length and the frequency. For these shortages, a fast method for repeats identification of repeat families via extension of consensus seed is proposed in this paper, which enables a rigorous definition of repeat boundaries and is based on the variant of BLAST algorithm. The known C.briggsae is used for testing the Repeat- Searcher. RepeatSearcher is compared with RECON, the most popular repeats identification algorithm, and the newly developed RepeatScout. The experimental results indicate that Repeat- Searcher has more accurate boundaries for each repeats, and the time of RepeatSearcher is reduced as compared with other methods with guaranteed accuracy.