SORS: Algorithms, software, and hardware accelerators for the next wave of genomic data

Abstract

In this talk, I will discuss how emerging fields such as pangenomics, pathogen surveillance, wastewater epidemiology, comparative genomics, and metagenomics are resulting in new waves of genomic data and applications. I will also discuss the various computational and storage challenges this data presents and how at Turakhia lab, we are using a combination of new algorithms, software, FPGA, GPU, and high-performance computing (HPC) solutions to address them. In pangenomics, we introduced PanMAN, a compact and unified data representation that integrates phylogeny, mutational history, genomic variation, and whole-genome alignments—making it the first of its kind. PanMAN was used to construct the largest pangenome for SARSCoV- 2 currently available, of over 8 million sequences, which requires only 366MB of disk space. This was enabled in part by TWILIGHT, our GPU-accelerated multiple sequence aligner that offers orders-ofmagnitude speedups and scales far beyond existing tools. For pathogen surveillance, we developed the UShER toolkit, which enabled real-time SARS-CoV-2 genomic surveillance and epidemiological research at a global scale during the COVID-19 pandemic, and has contributed to the designation of over 4,000 lineages. Building on UShER, we recently created WEPP, a novel HPC tool that significantly enhances the resolution and timeliness of wastewaterbased epidemiology, and is enabling powerful new applications. In comparative genomics, we developed ROADIES, an HPC software that fully automates accurate species tree inference from raw genome assemblies. ROADIES is transforming large-scale phylogenetic studies and is currently being used to analyze assemblies from the Vertebrate Genomes Project (VGP). Lastly, I will share my vision for how hardware accelerators can drive the next wave of innovation in bioinformatics, including some of our work based on high-level synthesis.