Authors

Fei He

Type

Text

Type

Dissertation

Advisor

Wu, Song | Zhu, Wei | Wang, Xuefeng | Jia, Jiangyong.

Date

2016-12-01

Keywords

Applied mathematics -- Bioinformatics

Department

Department of Applied Mathematics and Statistics

Language

en_US

Source

This work is sponsored by the Stony Brook University Graduate School in compliance with the requirements for completion of degree.

Identifier

http://hdl.handle.net/11401/77144

Publisher

The Graduate School, Stony Brook University: Stony Brook, NY.

Format

application/pdf

Abstract

Rapid and automated next generation sequencing (NGS) methods have emerged recently and significantly accelerated the research in biological and medical fields. The high-throughput NGS usually generates billions of shorter reads, which poses great bioinformatics challenges on extracting meaningful information from these massive data, one of which is de novo assembly. At the same time, the fast development of massive parallel processing (MPP) systems presents a substantial opportunity for processing larger datasets. Therefore, using supercomputer innovations on NGS research promises a good strategy; however, this application is not straightforward and requires new algorithms and parallel design for efficient implementations. In this thesis, we develop and present PPLAT, an integrated hierarchical multitasking parallel platform framework, and PPASSEM, a novel genome assembler built on PPLAT. PPLAT is designed for distributed storage and distributed processing of big data by enabling asynchronous computing and message passing, and provides a hybrid of multithreading- and MPI-based solution for MPP systems with simple APIs and great flexibility. We demonstrate the power of PPLAT to significantly reduce the coding and debugging complexity as well as facilitate high performance of derived parallel programs. PPASSEM is a novel application built on PPLAT, which employs the small-scale shared-memory multithreading and the large-scale distributed-memory parallelism using de Bruijn graph data structure for short–read sequences data. Our parallel platform has been tested on commodity computer clusters, based on both simulated and real data. Our results show that PPLAT can effectively handle billions of short reads (~500GB), and PPASSEM can generate accurate assembly constructs with much less time, compared with other well-known benchmark assembler like ABySS and PASHA. As new additions to the existing NGS toolbox, we expected that PPLAT and PPASSEM will greatly facilitate the future NGS-based research. | Rapid and automated next generation sequencing (NGS) methods have emerged recently and significantly accelerated the research in biological and medical fields. The high-throughput NGS usually generates billions of shorter reads, which poses great bioinformatics challenges on extracting meaningful information from these massive data, one of which is de novo assembly. At the same time, the fast development of massive parallel processing (MPP) systems presents a substantial opportunity for processing larger datasets. Therefore, using supercomputer innovations on NGS research promises a good strategy; however, this application is not straightforward and requires new algorithms and parallel design for efficient implementations. In this thesis, we develop and present PPLAT, an integrated hierarchical multitasking parallel platform framework, and PPASSEM, a novel genome assembler built on PPLAT. PPLAT is designed for distributed storage and distributed processing of big data by enabling asynchronous computing and message passing, and provides a hybrid of multithreading- and MPI-based solution for MPP systems with simple APIs and great flexibility. We demonstrate the power of PPLAT to significantly reduce the coding and debugging complexity as well as facilitate high performance of derived parallel programs. PPASSEM is a novel application built on PPLAT, which employs the small-scale shared-memory multithreading and the large-scale distributed-memory parallelism using de Bruijn graph data structure for short–read sequences data. Our parallel platform has been tested on commodity computer clusters, based on both simulated and real data. Our results show that PPLAT can effectively handle billions of short reads (~500GB), and PPASSEM can generate accurate assembly constructs with much less time, compared with other well-known benchmark assembler like ABySS and PASHA. As new additions to the existing NGS toolbox, we expected that PPLAT and PPASSEM will greatly facilitate the future NGS-based research. | 93 pages

Share

COinS
 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.