[tt] [GreyThumb] Reminder, Meeting Tonight: Chris Dwan on Genomics and High Performance Computing

[Eugen Leitl]

----- Forwarded message from Christopher Dwan <chris at dwan.org> -----

From: Christopher Dwan <chris at dwan.org>
Date: Fri, 5 Oct 2007 16:04:03 -0400
To: Grey Thumb Boston <society at greythumb.org>
Subject: [GreyThumb] Reminder, Meeting Tonight: Chris Dwan on Genomics and High Performance Computing
X-Mailer: Apple Mail (2.752.3)
Reply-To: Grey Thumb Boston <society at greythumb.org>



Sorry for the slow response on this:

I've put my slides here:  http://bioteam.net/Greythumb.pdf

>And, can you direct some questions to him on the role he sees for Blue
>Gene (the IBM computer) in HPC for genomics applications?

My opinion is that bioinformatics is not limited as much by computing  
hardware as it is by a combination of poorly formed questions and  
poorly implemented interfaces.  Once something is a stable enough  
biological question to be amenable to the sort of high powered  
simulation that a Blue Gene system does ... it tends to be rather  
quickly subsumed by changes in the experimental science.  There are a  
few examples of really hardcore simulation that revealed interesting  
biological truths ... but not very many.

That said, I mentioned the Blue Brain project in my talk.  Also,  
someone ported a workhorse program called "BLAST" to Blue Gene.

>Also, is he familiar with Cell/BE and if so, does he have examples of
>people using it in genomics research?

I know that the national labs (Los Alamos in particular) have built  
out clusters of Cells and are doing some nifty stuff with them.   
Again, the problem is in finding a decent problem where raw compute  
speed or architecture is the rate limiting factor.

Biology, it should be noted, is a discipline with millions upon  
millions of fairly arbitrary constants and relationships.  Unlike  
physics, we haven't managed to distill a small set of fundamental  
rules that lead rather predictably to the complexity that we see  
around us.  Instead, biology (and genomics, and medicine) tend to be  
focused on the accumulation of massive numbers of individual facts.   
We group those facts (and the skills required to derive them) into  
disciplines about the size and shape that a human mind can  
accommodate in a lifetime of study.

It is very, very rare for experimental physics to come up with a  
truly unexpected behavior of matter, of late.

It is very, very common for an experimental biologist to observe  
something totally novel that re-shapes thinking in the field.   
Happens all the time.

>And, the final question, would he be willing to watch and listen to
>Adam's Grey Thumb talk and explore ways that Adam's thoughts on
>evolutionary computing can be used in genomics research? Maybe Adam  
>has
>thoughts on this?

I would be happy to watch it, but google video failed me when I  
clicked the link.

Generally speaking, I'm not that big a fan of applying "biologically  
inspired" models back to research in biology.  I think that  
evolutionary algorithms, neural nets, and the like are quite powerful  
tools for finding unexpected patterns in large sets of data ... but  
they are really not good models at all of how living organisms  
work.   There are some cases where this is not true ... predator:prey  
relationships come to mind and particularly amenable to simulation.

That said, there are a lot of large sets of data available, and some  
of it is even well curated.  I'm sure that there is valuable insight  
to be gained from trolling through it.  I just warn against taking  
the overlapping terms too seriously.  Neural nets and neurology might  
share a latin root ... but they are very different disciplines.  It's  
almost equally unlikely that a neuroscientist will have good advice  
on converging a backprop. simulation ... as that an AI expert would  
have good advice on where to cut in the brain in order to stop a  
seizure.

Thanks again for the opportunity to talk!

-Chris Dwan



-----------------------------------------------------------
This message was sent via the Grey Thumb Boston mailing
list.  Visit http://www.greythumb.org/list for more info.
-----------------------------------------------------------

----- End forwarded message -----
-- 
Eugen* Leitl <a href="http://leitl.org">leitl</a> http://leitl.org
______________________________________________________________
ICBM: 48.07100, 11.36820 http://www.ativel.com http://postbiota.org
8B29F6BE: 099D 78BA 2FD3 B014 B08A  7779 75B0 2443 8B29 F6BE