Email updates

Keep up to date with the latest news and content from BMC Research Notes and BioMed Central.

Open Access Highly Accessed Technical Note

CloudAligner: A fast and full-featured MapReduce based tool for sequence mapping

Tung Nguyen1*, Weisong Shi1 and Douglas Ruden2

Author Affiliations

1 Computer Science Department, Wayne State University, US

2 Institute of Environmental Health Sciences, Wayne State University, US

For all author emails, please log on.

BMC Research Notes 2011, 4:171  doi:10.1186/1756-0500-4-171

Published: 6 June 2011

Abstract

Background

Research in genetics has developed rapidly recently due to the aid of next generation sequencing (NGS). However, massively-parallel NGS produces enormous amounts of data, which leads to storage, compatibility, scalability, and performance issues. The Cloud Computing and MapReduce framework, which utilizes hundreds or thousands of shared computers to map sequencing reads quickly and efficiently to reference genome sequences, appears to be a very promising solution for these issues. Consequently, it has been adopted by many organizations recently, and the initial results are very promising. However, since these are only initial steps toward this trend, the developed software does not provide adequate primary functions like bisulfite, pair-end mapping, etc., in on-site software such as RMAP or BS Seeker. In addition, existing MapReduce-based applications were not designed to process the long reads produced by the most recent second-generation and third-generation NGS instruments and, therefore, are inefficient. Last, it is difficult for a majority of biologists untrained in programming skills to use these tools because most were developed on Linux with a command line interface.

Results

To urge the trend of using Cloud technologies in genomics and prepare for advances in second- and third-generation DNA sequencing, we have built a Hadoop MapReduce-based application, CloudAligner, which achieves higher performance, covers most primary features, is more accurate, and has a user-friendly interface. It was also designed to be able to deal with long sequences. The performance gain of CloudAligner over Cloud-based counterparts (35 to 80%) mainly comes from the omission of the reduce phase. In comparison to local-based approaches, the performance gain of CloudAligner is from the partition and parallel processing of the huge reference genome as well as the reads. The source code of CloudAligner is available at http://cloudaligner.sourceforge.net/ webcite and its web version is at http://mine.cs.wayne.edu:8080/CloudAligner/. webcite

Conclusions

Our results show that CloudAligner is faster than CloudBurst, provides more accurate results than RMAP, and supports various input as well as output formats. In addition, with the web-based interface, it is easier to use than its counterparts.