DNA sequence alignment: From algorithm to systolic architecture

Biological sequence alignment is time-consuming tasks. This is due to its repetitive alignment matrix scores calculations. The SSEARCH35 is the well-known `software-only' sequence homology search tool. However, its sequential execution requires quadratic time complexity even when the software is executed in a sophisticated computer. Alternatively, sequence alignment is executed in hardware such as FPGA. FPGA allows for realization of systolic array (SA) architectures on its sea of logic gates. SA has been proven to address the aforementioned issues, whereby time required to compute alignment matrix scores in an n by n matrix significantly improved from n
²
to 2n-1. However, SA requires more hardware resources as the processing element (PE) is replicated to increase performance throughput. This paper presents an optimized and parameterizable PE architecture targeting for DNA sequence alignment with both affine and linear gap penalty score. The PE was designed and synthesized in the Quartus II design environment. The DNA sequence alignment core was implemented on ALTERA Cyclone II FPGA with device number EP2C70F896C6. The functionality of PE has been successfully tested and verified against the SSEARCH35 software. Results have shown that, the core able to operate more than 50 MHz when implemented in FPGA. With less than 1% area utilization/PE, the FPGA device can generate more than 200 PEs for single pass processing.