Accelerating pairwise sequence alignment on GPUs using the Wavefront Algorithm

Advances in genomics and sequencing technologies demand faster and more scalable analysis methods that can process longer sequences with higher accuracy. However, classical pairwise alignment methods, based on dynamic programming (DP), impose impractical computational requirements to align long and noisy sequences like those produced by PacBio, and Nanopore technologies. The recently proposed Wavefront Alignment (WFA) algorithm paves the way for more efficient alignment tools, improving time and memory complexity over previous methods. Notwithstanding the advantages of the WFA algorithm, modern high performance computing (HPC) platforms rely on accelerator-based architectures that exploit parallel computing resources to improve over classical computing CPUs. Hence, a GPU-enabled implementation of the WFA could exploit the hardware resources of modern GPUs and further accelerate sequence alignment in current genome analysis pipelines. This thesis presents two GPU-accelerated implementations based on the WFA for fast pairwise DNA sequence alignment: eWFA-GPU and WFA-GPU. Our first proposal, eWFA-GPU, computes the exact edit-distance alignment between two short sequences (up to a few thousand bases), taking full advantage of the massive parallel capabilities of modern GPUs. We propose a succinct representation of the alignment data that successfully reduces the overall amount of memory required, allowing the exploitation of the fast on-chip memory of a GPU. Our results show that eWFA-GPU outperforms by 3-9X the edit-distance WFA implementation running on a 20 core machine. Compared to other state-of-the-art tools computing the edit-distance, eWFA-GPU is up to 265X faster than CPU tools and up to 56 times faster than other GPU-enabled implementations. Our second contribution, the WFA-GPU tool, extends the work of eWFA-GPU to compute the exact gap-affine distance (i.e., a more general alignment problem) between arbitrary long sequences. In this work, we propose a CPU-GPU co-design capable of performing inter and intra-sequence parallel alignment of multiple sequences, combining a succinct WFA-data representation with an efficient GPU implementation. As a result, we demonstrate that our implementation outperforms the original WFA implementation between 1.5-7.7X times when computing the alignment path, and between 2.6-16X when computing only the alignment score. Moreover, compared to other state-of-the-art tools, the WFA-GPU is up to 26.7X faster than other GPU implementations and up to four orders of magnitude faster than other CPU implementations.