ACADEMIA

Pico Computing will demonstrate its hardware accelerated Burrows-Wheeler Aligner (BWA) at the SC12 conference, to be held on November 10-16, 2012 in Salt Lake City, UT. For high throughput sequencing operations, BWA can take up to 8 hours to complete on commodity CPUs. However, with Pico's hardware accelerated solution, this process has been reduced to less than 10 minutes resulting in a 480% improvement.

In the past ten years, major improvements in DNA analysis have been made. With each improvement, the amount of data needing to be processed dramatically increases. One of the first stages when analyzing sequenced data is to map the short pieces, known as short reads, to a reference genome. This inexact matching process is commonly performed on commodity CPU clusters using a software package called BWA. As advancements in next generation sequencing technology continue to occur, the amount of computational requirements also continue to increase. As a result, the overhead costs associated with managing, updating and adding additional nodes to process the increasing amount of data have become an alarming concern.

Pico is proud to announce the BWA algorithm as its newest addition to their bioinformatics library. The solution from Pico comprises 12 M-505 FPGA modules in their SC5 SuperCluster. Pico has accelerated BWA by pushing some of the application logic into the M-505s and parallelizing the memory accesses to BWA's index for increased memory bandwidth. Pico's solution draws only 120 Watts. Comparatively, in order for commodity CPUs to achieve the performance levels of 12 M-505s, a total of 12 quad-core CPUs are required, drawing approximately 1.5 kilowatts.

"With a low-memory alignment algorithm like BWA, we can push the reference genome index into the DRAM local to each FPGA to efficiently parallelize memory lookups. Integrating all software command line options with our SC5 running BWA, researchers and biologists can seamlessly transition their workflows to our accelerated FPGA solution" said Paul Draghicescu, Software Engineer at Pico Computing.