Presentation
Profile-Driven Banded Smith-Waterman acceleration for Short Read Alignment
DescriptionSmithWaterman is a popular DP algorithm utilized in alignment of massive and highly accurate short read datasets.
Its complexity requires drastic optimization techniques, such as hardware acceleration, heuristics and pre-filtering.
This work combines these approaches into a single powerful accelerated solution that leverages the low error profile of reads and employ the principles of Banded SmithWaterman, to highlight the value of creating accelerators customized to the accuracy requirements of the datasets. We prosose a multi-accelerator design that provisions both Banded and Full SmithWaterman components in order to meet the demands of the datasets in both high throughput and high acucuracy.
Its complexity requires drastic optimization techniques, such as hardware acceleration, heuristics and pre-filtering.
This work combines these approaches into a single powerful accelerated solution that leverages the low error profile of reads and employ the principles of Banded SmithWaterman, to highlight the value of creating accelerators customized to the accuracy requirements of the datasets. We prosose a multi-accelerator design that provisions both Banded and Full SmithWaterman components in order to meet the demands of the datasets in both high throughput and high acucuracy.
Event Type
Research Manuscript
TimeTuesday, July 11th1:40pm - 1:55pm PDT
Location3012, 3rd Floor
Design
SoC, Heterogeneous, and Reconfigurable Architectures