Events

Accelerating genomic computation to enable new discoveries

Genome sequencing data continues to rise faster than any other data type (80%/year), while transistor performance scaling has slowed down considerably (3%/year). New medical and comparative genomics applications have emerged that ensure there will never be “enough” sequencing. As genomic data overwhelms our compute capacity, domain-specific acceleration (DSA), i.e. using specialized hardware for accelerating a narrow domain (like genomics or machine learning), remains one of the few approaches in computer architecture to continue to scale compute performance and efficiency to enable the vast potential of genomic data.

In this talk, I will present our work on designing co-processors for two emerging, compute-intensive applications in genomics — long read assembly (Darwin) and whole-genome alignments (Darwin-WGA), together with the DSA and biology lessons learned from them. I will demonstrate how specialized hardware can provide massive speedup (1,000-10,000x) for genomic computations using a hardware-software co-design approach. Hardware can sometimes also facilitate new discoveries by enabling computations that are otherwise prohibitive in software. In the second half of my talk, I will show how we used whole genome alignments to make fascinating biological discoveries, including finding the molecular basis of convergent evolution and hundreds of gene losses in mammals. In the end, I will provide my long-term vision for hardware acceleration applied to emerging problems in genomics.