[BiO BB] Implementing HMM models in Hardware (FPGA)

Martin Gollery mgollery at unr.edu
Fri Sep 12 13:14:36 EDT 2003

Hi Venu,
  Yes, this idea makes sense. Anyone who uses InterProscan wishes it were 
faster! The current customer base includes all the big genome centers, 
universities, many biotechs and larger Pharmas. Other people have done projects 
in the field- Kestral at UCSC has been around for years, I don't know if it has 
been updated recently. Hokiegene out of Virginia is a newer system that shows 
great performance. There are several other commercial ventures that are in 
stealth mode right now.
   HMM's really lend themselves well to acceleration of this type. Perhaps you 
could focus on accelerating SAM instead of HMMer, as many people prefer it and 
TimeLogic, Compugen and Paracel do not accelerate SAM. Also, nobody (to my 
knowledge) has accelerated ClustalW, so this might be something else to 

Best of luck! Sounds like an interesting project!


3. Quoting Venu Kode <vkode78 at yahoo.com>:

> Hello everyone,
> I am graduate student in computer engineering doing my thesis in the
> following topic:
> "Matching Protein sequences with HMM models in FPGAs ( Field Programmable
> Gate Arrays ) using Run Time Reconfiguration"
> If you are familiar with Decypher tool from TimeLogic, my work involves
> something similar to that.  Decypher is a very comprehensive and expensive
> tool and sure it does deliver excellent performance compared to the software
> tools such as HMMR, SAM or the other tools. My work falls in between the
> software solution and that of Decypher.
> Specifically, when developed my tool would be considerably faster than that
> of the software tools currently available and also be affordable at a
> fraction of the cost of using Decypher. 
> I have just started digging into Bioinformatics and have read quite a number
> of papers and all, but I am still a little confused and would like any
> comments or suggestions from you:
> 1) Does my tool make any sense at all? 
> 2) What is the current customer base like for this technology? 
> 3) What sort of companies do the work of matching protein/DNA sequences with
> existing models? 
> 4) Is there a need for a less comprehensive and less expensive tool as
> opposed to Decypher for customers who want to get it done a lot less cheaper
> but wouldnt mind the extra penalty in performace ( ofcourse will be very much
> better than that of the software searching)
> 5) Any comments, questions, suggestions?
> 6) Any pointers for me in terms of websites or resources.
> I would very much appreciate your comments
> Kode
> ---------------------------------
> Do you Yahoo!?
> Yahoo! SiteBuilder - Free, easy-to-use web site design software

Martin Gollery
Associate Director of Bioinformatics
University of Nevada at Reno
Dept. of Biochemistry / MS330
New phone number! 775-784-7042

This mail sent through https://webmail.unr.edu

More information about the BBB mailing list