The recent breakthroughs in next-generation sequencing technologies, such as those of Roche 454,Illumina/Solexa, and ABI SOLID, have dramatically reduced the cost of producing short reads of the genome of new species. The huge volume of reads, along with short read length, high coverage, and sequencing errors, poses a great challenge to de novo genome assembly. However, the paired-end information provides a new solution to these problems. In this paper, we review and compare some current assembly tools, including Newbler, CAP3, Velvet,SOAPdenovo, AllPaths, Abyss, IDBA, PE-Assembly, and Telescoper. In general, we compare the seed extension and graph-based methods that use the overlap/lapout/consensus approach and the de Bruijn graph approach for assembly. At the end of the paper, we summarize these methods and discuss the future directions of genome assembly.
Yiming HeZhen ZhangXiaoqing PengFangxiang WuJianxin Wang
Evidence shows that biological systems are composed of separable functional modules. Identifying protein complexes is essential for understanding the principles of cellular functions. Many methods have been proposed to mine protein complexes from protein-protein interaction networks. However, the performances of these algorithms are not good enough since the protein-protein interactions detected from experiments are not complete and have noise. This paper presents an analysis of the topological properties of protein complexes to show that although proteins from the same complex are more highly connected than proteins from different complexes, many protein complexes are not very dense (density 0.8). A method is then given to mine protein complexes that are relatively dense (density 0.4). In the first step, a topology property is used to identify proteins that are probably in a same complex. Then, a possible boundary is calculated based on a minimum vertex cut for the protein complex. The final complex is formed by the proteins within the boundary. The method is validated on a yeast protein-protein interaction network. The results show that this method has better performance in terms of sensitivity and specificity compared with other methods. The functional consistency is also good.
