Bilkent University
Department of Computer Engineering




Zülal Bingöl
PhD Student
(Supervisor: Assoc. Prof. Can Alkan )
Computer Engineering Department
Bilkent University

Abstract: High throughput sequencing (HTS) advances facilitate the fast production of short DNA fragments (reads) in numerous amounts. Long-read sequencing technologies have been accompanying short-read sequencers with improved accuracy and cost-efficiency in the past few years, revolutionizing biological research. Unlike short reads, long reads are better for genome assembly and easier to map to regions with high repeat content. Because of both their length and higher error rates, long read mapping algorithms are different than that from short reads. Although there have been considerable improvements in tools to align long-reads, read mapping using only the main processor (CPU) is still slow and a bottleneck in genome analysis. In this sense, general-purpose GPU accelerators are becoming more prevalent as they offer an alternative platform for obtaining benefits from performing compute-intensive work with highly parallel and independent millions of threads for boosting performance. Our goal is to design an end-to-end mapping tool for long reads in CPU-GPU heterogeneous systems in this project. Our initial plan is to release each step of the mapping pipeline as a separate unit in a GPU accelerator and develop novel algorithms as performing them in GPU provides performance gains.


DATE: 18 April 2022, Wednesday @ 15:30 Zoom