Summary

Genome analysis is key for diagnosing genetic diseases and developing personalized cancer treatments. However, the rapid growth of genomic data is outpacing current computing power, slowing down medical research and delaying important diagnoses. Traditional computer systems can’t handle this data explosion due to limitations like the Von Neumann bottleneck, i.e., the computing performance is constrained by the data movement between the memory and the CPU. To address this, recent innovations in computing propose performing computation in memory or closer to the memory. GenomICs adopts these novel data-centric computing paradigms to accelerate genome analysis, particularly focusing on the DNA long-read analysis pipelines. By reducing the back and forth movement of gigabytes of data on the memory bus, genomICs delivers a significant speedup in the analysis time while simultaneously reducing the energy consumption.

In addition to the novel hardware, genomICs also employ a novel software-stack to effectively utilise the exploit the underlying hardware. The software-stack preciely identifies the most computationally demanding tasks in the analysis process and optimizes them to accelerate the entire process. It also readily integrates existing analysis methods without requiring changes to the source codes.