Environment for accelerated bioinformatics applications

May 11, 2012 at 12:59 PM

I (and many others) have been working on accelerating bioinformatics applications with FPGAs with an overlapping community working with GPUs.  Applications include the usual suspects and more: Smith-Waterman and related DP-based algorithms, BLAST (P,N,Psi, etc.), HMMer, Tandem Repeats, ClustalW, Bowtie, etc.  While really impressive results have been obtained I think that there has been not nearly the distribution/utilization that we would like.

Would .NET Bio be an appropriate environment to integrating and disseminating accelerated applications?

-Martin Herbordt,

May 11, 2012 at 3:04 PM

Hi Martin,

If you are doing bioinformatics on the Windows platform, yes absolutely it could make sense to use .NET Bio, and we would be happy to welcome you and your colleagues into the community. The first step would be for you to take a closer look at the .NET Bio project, to see if it has the capabilities you need, so feel free to download, and also look at the documentation. I'm also happy to answer any questions you may have.

What GPU acceleration techniology are you currently using, and how many apps have you accelerated? Do you have a project page, or somewhere we can learn more about your work?



May 11, 2012 at 5:34 PM

To Martin:

1, Seems to me, that anything you have done, need to be published either on Bioinformatics; BMC Bioinformatics; better to be Nature Method etc. then there will be so called bioinformaticians even start to look at your work. When they actually did some real work with your work, and got it published, you may expect more to follow. This could be very long process, that given the moore's law, your speedup of 10x, kind of reduced to 4x because the CPU is faster or have more cores.

2, FPGA is even rare than GPU, and both of these are special hardware, to my knowledge there is not yet Free Hardware movement, so how does these people justify the budget, to their boss? And if not talking about one piece but an infrastructure, e.g. an FPGA/GPU farm, then in many places this will escalate into politics.

3, Most research groups, do not have BIG DATA, he who has BIG DATA, has BIG MONEY to buy whatever he likes, and normally able to have a group of programmers creating own tools, even they are the sole user. Just look at all these big sequencing centres over the world.

4, Windows is so rare in this bioinfo world, so ... ...

p.s. I personally is looking at Intel's many core, would like to know your opinion on it, the Many Core v.s. FPGA v.s. GPU