A Novel Associative Memory Based Architecture for Sequence Alignment

• Published in: 2016 IEEE International Parallel and Distributed Processing Symposium Workshops (IPDPSW) pp. 473 - 478
• Main Authors: Mirzaei, M. Ali, Crescioli, Francesco, Viret, Sebastien, Tromeur, William, Calderini, Giovanni, Marchiori, Giovanni, Baulieu, Guillaume, Galbit, Geoffrey
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.05.2016
• Subjects: AMchip; Associative memory; Computer aided manufacturing; Computer architecture; Field programmable gate arrays; Hamming Weight; Hardware; Microprocessors; NCBI BLAST; sequence alignment; Xilinx FPGA
• DOI: 10.1109/IPDPSW.2016.21

This paper presents a novel hardware architecture based on Associative Memory technology for sequence alignment. The Associative Memory chip (AMchip) architecture employs a huge amount of parallelization to perform real time combinatorial pattern matching. It can be used to perform very fast searches over a large database using hamming weight as a similarity metric. In the presented hardware platform the AMchip communicates with a Xilinx Zynq ARM CPU + FPGA through high-speed (2.4~Gbps) serial links enabling seamless integration between the software and the hardware acceleration. The result shows that low score sections of a database sequence are eliminated rapidly using this architecture which in turn leads to significant speed up. The outcome of a preliminary study on a selected reference protein database shows that all the matches between query and database sequences found by NCBI-BLAST can be successfully found also by the Associative Memory based algorithm.