Abstract
For the fast construction of multiple sequence alignment of large dataset, we improved MAFFT to use with GPGPU. We focused on accelerating Distance Matrix calculation, because it consumes 88% (15,000 sequences, 804 aa) of processing time without order dependency. For this purpose, we applied GPU as a good parallel processing capability. In addition, the time complexity affected by the sequence length was reduced by optimizing the calculation order. As a result, we obtained 4 times speed-up on 16,000 hemagglutinin sequences of the Influenza virus.