[Pipet Devel] Check out this URL

[Harry Mangalam]

Hi Again,

   Well, this wasn;t part of my original interest in this group and may be
well-suited for it, but let me describe one of the things I'm
working on (partially supported by National Center for Genomic Resources
(NCGR, out of Santa Fe, NM) in support of a yeast genomics project at UC
Irvine.

The UCI group has gotten an Affymetrix Genechip machine and is busy
subjecting yeast to various stresses, generating whole-genome datasets for
time points along this stress.  I'm building a relational database with a
web interface that will suck up those datasets (and be amenable to accepting
data from other such gene expression studies) and allow it to be queried on
various params, as well as subjecting the returned values to various
statistical analyses with the stats language 'R' (a clone of S/SPlus), using
gnuplots for the simple outputs, VRML for complex viz's.  

Because the size of the datasets are so large (6k orfs x 4 timepoints, plus
associated pointers, descriptors, images, etc) and the number of them is
going to be pretty big, I'm using mysql as a prototyping system, with perl
glue, talking thru Apache/FASTCGI, replacing the perl with C as I
identify bottlenecks. There will be a generic interface to commandline apps
(other clustering routines, tacg, clustalw, blast, etc, so that it can
become pretty extensible.  NCGR may rewrite it at commercial
quality to support their plant genomics project, but I get to do the fun
part...

I hadn't considered it, but you bring up the possibility of using such a
system as a collaboratory by making the analyses persistent in some way,
either as paths thru an analysis or the analysis itself (altho that would
get very large very fast) so that they might be re-used or extended by
others interested in the topic.  Or maybe just the paths thru an analysis
would be an important resource - if I could somehow record the 'analysis
track' that users took, I could identify, then automate them so that the
whole pathway could be boiled down to a button.

This is WELL off the LOCI topic, but perhaps the 2 could be designed to
communicate at some level.
As I said, it was never the intent for the above-described project to use
LOCI, but if they can be made to better co-exist so much the better.

Cheers
Harry

On Fri, 26 Feb 1999 david.lapointe at umassmed.edu wrote:

/Yeah, I realized that just after I sent that message.
/
/$2M ? Seems like a lot but if you've invested $20 million( or more)
/in sequencing hardware what's $2M to make it work?
/
/Are you talking about Collaboratories? That is an interesting concept.
/
/
/David
/
/David Lapointe
/Manager - Research Computing Services
/UMass Medical School
/Worcester, MA 01655
/508/856-5141
/
/
/> -----Original Message-----
/> From: J.W. Bizzaro [mailto:bizzaro at bc.edu]
/> Sent: Friday, February 26, 1999 5:06 PM
/> To: tulip-list at busboy.sped.ukans.edu
/> Subject: Re: [Pipet Devel] Check out this URL
/>
/>
/> Jeeeez!  Does the concept seem a little familiar?  BTW, this
/> is the same company
/> from which I got the first pics.
/>
/> You know, I have been thinking seriously about taking Loci
/> one step further and  making it a system for Internet-wide research
/collaboratives,
/> between loosely affiliated people.  It's something I still have to clear
/with
/> Ken Marx, but I was thinking that we, The BIC Group, could use Loci to
/> collaborate on some "open" research projects, making an "open laboratory"
/that
/> treats scientific research like a GNU software project.  Any thoughts?
/>
/>
/> Jeff
/> bizzaro at bc.edu
/>
/

Cheers,
Harry

Harry J Mangalam -- (949) 856 2899 -- mangalam at home.com