Bioinformatics—Leveraging the Power of FPGAs
In looking into the bioinformatics area, I was surprised by the lack of standards. There is a great deal of homebrew software. BLAST (Basic Local Alignment Search Tool) and SAGE (Serial Analysis of Gene Expression are about the only standards available. BLAST takes two gene sequences (i.e. string sequence 1 -- ATTTACCTGGT and string sequence 2 --ATTTAGGTCCT) and compares them to see if they are the same or similar. SAGE analyzes gene expression patterns with digital analysis. After that, a host of small company niche players and university developed tools comprise the landscape. The reason for this is the nature of government funding. A number of companies invested heavily into bioinformatics tools and databases only to see the US Government come along and provide similar tools at little or no charge. This wiped out the market for developing tools and opened the door for a host of custom-developed software.
Most university-developed software tools gain little following outside the lab where it is developed and without a revenue stream, there’s little incentive to provide support. The one exception to this is MEGA (Molecular Evolutionary Genetics Analysis) which performs gene sequencing and alignment and web-based data mining. Developed by Dr. Sudhir Kumar of Arizona State University, the software has over 50,000 downloads making it one of the more popular tools for bioinformatics research.
FPGA’s are increasingly used in gene sequencing because it speeds up the pattern matching algorithm by breaking the pattern into sub-strings and matching on each part simultaneously.
A number of companies develop FPGA based solutions for bioinformatics work including Impulse, Timelogic, and Mitrion based in Los Angeles who is developing a BLAST application to run on it’s FPGA-based Virtual Processor. They took the open source version of NCBI’s BLAST algorithm and rewrote portions of it to run in C-code on their processor. They claim 10-100 times performance improvement. This seems to be a common method – take standard, open source algorithms and modify them to work on an FPGA.
Xilinx, the leading vendor of FPGA technology, offers this paper on an FPGA-architecture.
I believe we’ll FPGA’s become more common in bioinformatics applications.