Mercogliano, Stefano (2024) Mendel, A Hardware Co-designed Energy Efficient and High-Performance Heuristic Pipeline for Large-scale Genome Analysis. [Tesi di dottorato]

Documento PDF mercogliano_stefano_37.pdf Visibile a [TBR] Amministratori dell'archivio Download (13MB) | Richiedi una copia

Abstract

High-throughput modern DNA sequencing has transformed our understanding of life at the molecular level, providing unprecedented power to decode the genetic blueprint of organisms on a large scale. Since the advent of DNA sequencing in 1977, the field has made remarkable progress. Today, third-generation sequencing platforms offer real-time sequencing of long DNA molecules, vastly enhancing the throughput of genomic data. These advancements have paved the way for critical applications across various domains, including personalized medicine, outbreak surveillance, evolutionary studies, and criminal forensics. However, the rapid increase in sequencing throughput has not been met with a corresponding rise in computational capacity. The volume of genomic data is growing faster than Moore’s Law, leading to petabytes of data generated annually and pushing towards exabyte-scale datasets in the near future. This massive data explosion places a significant strain on current computational infrastructures, which struggle to process such vast amounts of information without incurring unsustainable economic and energy costs. This challenge is particularly significant in read mapping, the core task in genome analysis that involves aligning short DNA fragments, or reads, to a reference genome. The computational intensity of read mapping has been exacerbated by the data deluge, necessitating new algorithmic and technological breakthroughs. In this work, we present Mendel, a heuristic-based software/hardware architecture designed to address these challenges. Mendel significantly reduces the amount of data processed during read mapping, thereby lowering both execution time and energy consumption on both software and high-performance hardware platforms. Our evaluation demonstrates that Mendel achieves orders-of-magnitude improvements in energy efficiency and speed compared to state-of-the-art solutions, positioning it as a potential milestone in the era of exabyte-scale genomics.

Downloads

Actions (login required)

Modifica documento