GPU Acceleration for Availability Scoring of Short Constituent Amino Acid Sequences

Frequency-based availability analysis for short constituent amino acid sequences provides us with novel possibilities for genome informatics researches specially such as homology search and motif detection. However, it requires very long computation time because of the huge data size and combinatorial explosion. In order to relax the computation time problem, this paper considers the GPU acceleration for the availability scoring of short constituent amino acid sequences. We aim drastic speed up by implementing the most time-consuming process in the availability scoring, the frequency computation for short constituent amino acid sequences in proteins as GPU functions. The experimental evaluation with real data shows that our GPU acceleration achieves 48 times speedup, compared to the CPU program.