Join us for our upcoming SAFARI Live Seminar:
Title: Software/Hardware Co-design and Dataflow acceleration for Short Read Alignment
DNA read alignment is an integral part of genome study, which has been revolutionized thanks to the growth of Next Generation Sequencing (NGS) technologies. The inherent computational intensity of string-matching algorithms such as Smith-Waterman (SmW) and the vast amount of NGS input data, create a bottleneck in the alignment step and stress current computing power. Literature is rich with optimization efforts that target read alignment and hardware acceleration in particular has been extensively leveraged to alleviate the bottleneck.
This talk presents two efficient software-hardware co-design architectures for the acceleration of short read alignment of NGS data. The first approach presents a dataflow accelerated system for short read alignment, that takes into account the implications of integrating the accelerator into a real system. It presents GANDAFL, a novel genome alignment dataflow architecture for Smith-Waterman Matrix-fill and Traceback stages to perform high throughput short-read alignment on Next Generation Sequencing data. The accelerator is then integrated into widely-used Bowtie2 aligner that has been restructured to implement an aggregation-batching strategy that minimizes data transfer and call overheads between the host and the accelerator. The second approach bridges hardware acceleration and pre-filtering techniques for alignment optimization and combines them into a data-driven approach that further optimizes the performance. Extensive profiling of genomic datasets reveals low edit thresholds that can be leveraged by Banded SmithWaterman to create resource-efficient accelerators that are customized to the edit profile of the input. We therefore propose a dataset-specific multi-dataflow design that leverages the profile-driven edit limits and resource-efficient Banded SmithWaterman accelerators to enable the provision of more accelerated units on the same device and perform read alignment at high throughput without accuracy loss.
Konstantina Koliogeorgi received her Ph.D. degree in Electrical and Computer Engineering in 2023 at National Technical University of Athens (NTUA) advised by Prof. Dimitrios Soudris. She is currently a Postdoctoral Researcher at the Microprocessors and Digital Systems Laboratory in NTUA. Her research interests lie in the field of computer systems and architecture, heterogeneous computing and hardware acceleration. Her research has focused on hardware-software co-design, efficient high level synthesis optimization and design space exploration, targeting mainly genome analysis applications.