An FPGA-based high speed DNA sequencer for early stage breast cancer diagnosis

Pairwise Deoxyribonucleic Acid (DNA) sequence alignment searches for regions of similarity by comparing each DNA nucleotide (NT) between a newly discovered (unknown or query) sequence against known sequences (subject sequence). Due to exponential growth of biological database and the time-consuming dynamic programming (DP)-based sequence alignment algorithms, researches on Field Programmable Gate Array (FPGA)-based accelerators have been extensively reported. The use of systolic array (SA)-based architecture in DNA sequence alignment has proven to be one of the best and most efficient ways to get alignment results in realistic time. However, aligning actual lengthy DNA sequences for a targeted sequence (which is of hundreds, if not thousands of DNA NTs) still requires longer computation time. Therefore, in this research, the DP-based global and local sequence alignment algorithms with affine gap penalty are optimized to exploit parallelism and pipelining offered by the SA-based processor architecture. A novel SA-based DNA sequencer architecture was designed, simulated, and synthesized on Virtex-6 FPGA with device number XC6VLX75T. The processing element (PE), which is the building block of the SA architecture, was designed using Verilog Hardware Description Language (HDL) and it implemented both the original and optimized sequence alignment algorithms (global and local) for comparisons and analysis. Results have shown that the PE area utilization of the proposed optimized algorithms reduced by 77% and 78% for the global and local algorithms respectively. This has enabled more PE duplications in the pipelined SA architecture to enable more parallel computations. In term of computational frequency, the optimized global algorithm PE architecture is faster in computation time by 40% (from 447 MHz to 750 MHz) as compared to the original global algorithm PE architecture. In the case of the local alignment algorithm, the optimized PE architecture has faster computation time by 17% (from 435 MHz to 527 MHz) compared to the original local algorithm PE architecture. For the targeted FPGA device, the optimized global and local DNA sequencer architecture is able to align DNA sequences with query length up to 994 NTs with 579.25 Giga Cell Updates Per Second (GCUPS) and 670 NTs with 216.00 GCUPS respectively. For actual disease detection, DNA sequence with highly mutated region of Breast Cancer susceptibility genes Type 1 (BRCA1) exon 11 DNA sequence has been identified and developed as a query sequence. The computation time of the proposed query sequence has shown improvement from 465 ms to 1.578 ns when ran on the optimized global algorithm DNA sequencer as compared to the (MATrix LABoratory) MATLAB computation time, whereas for the optimized local algorithm, the DNA sequencer is faster,(from 642 ms to 2.636 ns as compared to the MATLAB computation time.