| Summary: | 碩士 === 長庚大學 === 資訊工程學系 === 100 === Sequence alignment has been a very important issue in computational biology filed. The accuracy by Smith-Waterman sequence alignment algorithm is better than that of other sequence alignment algorithms. Unfortunately, the time complexity of it is high because it is done based on the dynamic programming methods to solve the problem. With more and more sequences generated in the sequence databases, this result makes the sequence alignment to spend more time. In recent years, many research works were proposed to implement the Smith-Waterman sequence alignment by CUDA and then got good results. We combine the CPU and CUDA computing capabilities to accelerate the Smith-Waterman sequence alignment. We calculate the frequency distance between the query sequence and database sequences in the CPU. The frequency distance can be screened for sequence alignment to achieve the purpose of reducing the number of alignments. We use the frequency distance removed sequences from the databases that the low similarity with the query sequence. In GPU, we implement the Smith-Waterman sequence alignment in the NVIDIA Tesla C2050 and enhance the speed of about 9 to 10 times based on our results. Finally, we combine the frequency distance and CUDA-SW, and the highest speed can be raised to 80 to 90 times.
|
|---|
