Pairwise Sequence Alignment Method for Distributed Shared Memory Systems

One of the initial key steps of the multiple sequence alignment problem is the pairwise alignment of all pairs of genomic sequences involved. With the increased requirements to align several thousand sequences, it is necessary to find efficient new ways to align as many pairs of sequences as possible. Traditional sequential algorithms are limited by their memory and processing capabilities while parallel implementations running over clusters are able to process considerably more sequences. Nowadays, computer systems are capable of running several processing threads using a shared memory model, which allows us to combine it with distributed memory model. This paper presents a parallel pairwise aligner based on Smith-Waterman capable of processing large numbers of sequences with a small memory footprint. Our implementation is based on the use of a message-passing library such as MPI combined with a threading library, such as pthreads. Our experimentation shows the gain in efficiency for processing different numbers of sequences with different numbers of threads.