We have previously released SWIPE for performing very rapid and accurate searches in biological sequence databases based a highly parallelised implementation of the Smith-Waterman optimal local alignment algorithm. Parallelisation using both SIMD, threads and MPI is exploited to achieve high performance. The recent Haswell processors released by Intel allow even better SIMD parallelisation using the 256-bit wide AVX2 extensions. Parts of the alignment methods in SWIPE have been reused in SWARM for clustering of DNA sequences from meta-genomic studies. The core alignment methods could also be valuable in other bioinformatics tools that are depended on effective sequence alignment. A library with an API that allows diverse tools to easily and effectively use the alignment functions provided would be very valuable.
This project involves designing a suitable API that allows a wide range of tools to use the alignment functions effectively. Furthermore, the API and library should be implemented based on the existing code base in SWIPE. Preferably, a version using the 256-bit wide AVX2 extensions should be implemented. The results must be evaluated.
The project is suitable for anyone interested in bioinformatics with some programming experience, preferably with parallelisation in one form or another.
Supervisor: Torbjørn Rognes (BMI/IFI)