[BiO BB] (no subject)

[Val Bykoski]

Hi Indraneel,
    Thanx for very informative comments and ideas.
    You say:
>I believe the concepts you point out are at present fairly complicated
> to represent and visualise using current software. Remember that most of
> current software was generated not to study the unknown diversity but to
> store man made discrete units (eg financial data). Possibly a giant step
> is required to produce such things and since I cannot at present see any
> necessity for it in the commercial world (since a .com will always want
> quick results), such a system will most probably be an open source
> development.
    Agree.  Very much so... Still, with the increasingly available (expression)
data,
the homology/analogy data/text processing is simply not enough, it
creates an ever increasing gap between the data and the biology context,
and thus makes drug discovery ineffective/unreliable.
Molecular details of cell mechanisms are indeed a highly complex stuff,
and it will take time to clarify the details.  Meanwhile, the coarse-grain
cell models with the rich (sequence-based) stimulus-response logic and
infrastructure might be a really good starting point in designing the
next-generation data-driven (and biochemistry/pathway-driven)
cell models that  incrementally refine cell structure and functions as soon
as more and more data is available.  With huge computational power and
storage available and powerful OO languages (eg, Python), design and
implementation of data-driven cell models seems to be quite practical now.

> The "parts" (that we study now) indeed make up the "whole", but we also
> need to remember, maybe according to "chaos" (or the gaia hypothesis?),
> that the whole is something more than just the parts.
    Totally agree.  That's what i call the cell infrastructure, and i believe
that existing kinetic cell models need to be complemented with an infrastructure
borrowed from physics and chemistry of (soft) condensed state such as
"aperiodic" crystals (with space-variant subunits) that allows for cell-wide
signaling, update, genome restructuring, etc

Open source  approach sounds great to me...
i think, though, that NIH/NSF would certainly be interested in this kind of
explorative research and development, and with enough interest/input from the
bioinformatics community i'd be willing to explore the proposal writing
opportunities...

my best,
/val
vbykosky at galileo.eng.uml.edu
val at vtek.com

----- Original Message -----
From: "Indraneel Majumdar" <indraneel at indialine.org>
To: <bio_bulletin_board at bioinformatics.org>
Sent: Tuesday, September 25, 2001 11:58 PM
Subject: Re: [BiO BB] (no subject)


> Hi Val,
>
> Sorry for the very late followup. After your mail I realised that truly
> we are basing most of our bioinfo ideas on things like homology/analogy
> rather than in terms of biochemical pathways, protein function prediction
> in terms of cellular requirements and such things (something that I
> wanted to do when I first entered this field, but lost sight of and
> forgot very soon).
> ... skipped
>
> On Sat, Sep 15, 2001 at 09:41:34PM -0400, Val Bykoski wrote:
> > Hi Indraneel,
> >     Thanx for your comments.
> >     I refer to the cell as a biological cell, and my point,
> >     to put it a little differently, was that data has to work
> >     inside a (realistic) cell model, not to sit passively in databases.
> >     Data grouping, visualization, etc. is better than nothing,
> >     but still there is a huge gap between data (nicely grouped,
> >     visualized, etc) and their biological significance or context.
> >     I guess my point is that if you integrate data into a realistic
> >     cell model, the model is both a predictive framework and
> >     a (bio)context-sensitive database for the data.  Then,
> >     data is hidden and work inside the model for predictions,
> >     and does not require grouping, visualization, etc.
> >     Sure, the "realistic cell model" is a hard nut...but may be
> >     worth of efforts, in particular, for a data-driven model.
> > cheers, val
> > val at vtek.com
> >
>
> --
> http://www.indialine.org/indraneel/
>
> _______________________________________________
> BiO_Bulletin_Board maillist  -  BiO_Bulletin_Board at bioinformatics.org
> http://bioinformatics.org/mailman/listinfo/bio_bulletin_board
>