Comparative analysis of protein alignment algorithms in parallel environment using CUDA

Abstract

In bioinformatics to identify evolutionary relationships two sequences are matched to find similarity. Smith Waterman, a dynamic algorithm, is a common choice to carry out this alignment process. However, with the exponential growth of protein databases this algorithm’s time complexity increases. The demand of bioinformatics for their tasks to speed up is very high. Even a slight seed up in computation would be very helpful in the field of bioinformatics. Thus, for a lot of the scientists this algorithm might not be the first choice. In today’s world the most popular and used bioinformatics tool is the BLAST (Basic Local Alignment Tool). BLAST, similar to Smith Waterman algorithm, is an alignment algorithm for scanning proteins from protein databases. This thesis analyzes both the algorithms in a parallel environment with the help of NVIDIA GPU. For our experiments we utilized a GeForce GTX 660 NVIDIA GPU to execute both the algorithms. Experimental results show that BLAST is on average is 2.5 times faster than Smith-Waterman.