[AG-TECH] virtual seminargenomics & bioinformatics: December 18 noon-1 pm ET

[Willy Valdivia Granda]

Virtual Seminar on Genomics and Bioinformatics
www.virtualgenomics.org

Thursday, December 18, 2003, Noon - 1PM Eastern time

Val Bykoski
Boston University

Understanding Cell Dynamics via Data Mining

A critical overview of cellular regulatory and expression mechanisms is
given with the focus on understanding cell dynamics based on direct use
of (microarray and other) data rather than various ad hoc models.
Data-driven methods to build the next-generation cell models are presented
and
discussed as well as the use of such models to simulate cell response to
potential
drugs, small molecules effects, etc.

A proposed approach introduces a generic framework to be customized by
data-driven training to describe a specific cellular structure and
functionality. That makes data to "talk" immediately inside the model
rather than to be visualized for a human expert to look at the data and make
a
judgment In the process of customization, an online cell model gets
built. Conceptually, the framework describes a cell as an aperiodic crystal
with a restructurable unit. Like in any crystal, signals can propa-gate from
cell periphery to nucleus and back as organized in space and time wave
fronts.

The signals may interfere and create patterns inside a cell. The
patterns from repeating signals get spatially overlaped providing a
generalization capabili-ty and a reinforcing effect. The microarray data can
be used to
build a specific cell model similarly to training a neural network. Such
a model can then be incrementally refined with new data, when available,
which makes it a persistent container (database) for the data with
prediction
capabilities - it generates a context associated with new data. Data for
yeast (S. cerevisiae) are used to demonstrate the approach, and the
results are presented and discussed.

***********
Dr. Val Bykoski has MS in Bioorganic Chemistry from Moscow Technology
University and PhD in Physics and Applied Math from Russian Academy of
Sciences. He was with the Institute for Chemical Physics, Moscow,
studying computationally quantum effects in chemical and biochemical
reactions At
the National Institute for Control Problems, Moscow, he was doing research
in mechanisms of memory and control in engineering and biological systems.
He was a Royal Society, London, Guest Professor lecturing at six UK
Universities and doing research in content-driven memory, and
demonstrated for the first time a holophone, a content-addressed memory
device for
pattern sequences (a possible brain memory model). He was awarded the
International Norbert Wiener Prize for his research in control and
memory mechanisms. In the US, he was with Boston University as a Faculty
Research Professor and was teaching at University of Massachusetts at Lowell
(CS
and EE Departments) and University of New Hampshire. He was a Consultant for
Raytheon, Xerox, and EMC, and again a Research Faculty at Boston