The project entitled high performance computation for short read alignment investigated high performance computational techniques for the analysis of ribosomal RNA, which is the mechanism that cells use to translate an organism's DNA into protein. Next generation sequencing techniques are enabling the capture of vast amounts of data on the ribosomal RNA characteristics of cells in varying conditions. However reads for such RNA fragments are smaller than those typically encountered in sequencing projects, hence most alignment algorithms are optimised for longer reads.
"The SHAPE project has been a very successful collaboration between NSilico and the PRACE partners involved. NSilico has benefited from domain expertise from PRACE in first identifying a bioinformatics codebase with real potential to be deployed on cutting-edge many-core hardware. It has also since gained invaluable insights into the optimisation and parallelisation work involved in porting the code to the Intel Xeon Phi. Next steps are already being discussed on testing and deployment of the code with the release of the next generation 'Knights Landing' hardware, as well as potential incorporation into NSilico's in-house bioinformatics pipelines", stated Paul Walsh of NSilico.
The project team adopted two approaches to optimise and parallelise the SSW library, first using modern SIMD intrinsics and the second using OpenMP. The OpenMP parallelisation work has led to a code that shows good parallel performance results on standard x86 processors and promising results for Xeon Phi many-core hardware. While the resulting SSW library achieves expectedly limited performance gains on the current generation of the Xeon Phi, it has been re-factored in a way to readily take advantage of the next generation of hardware such as Xeon Phi 'Landing' with upcoming AVX 512 features.
The results of the project were presented during the SHAPE parallel track of PRACEdays14 .