## Brock Institute for Scientific Computing |
||

Brock University,
500 Glenridge Avenue, St.Catharines,
Ontario, Canada L2S 3A1 (905)688-5550 |

The Brock Institute for Scientific Computation (BISC) was established this year to promote collaborative relations between computationally-based research groups within Brock University, in addition to collaborations with faculty at other Universities (presently McMaster) and industry. We have formed interdisciplinary teams of researchers with expertise in information science, biotechnology, computational biology, mathematics, and physics, experts in the use of numerical analysis, mathematical modelling, computer simulation, and artificial intelligence, to tackle major research problems demanding innovative solutions. These have practical technological implications across a broad spectrum of important Canadian industries.

The current director of BISC is Prof. Stuart M. Rothstein.

RESEARCH INTERESTS

Computational Physics [insert link here]

S. Bose

H. Fuks

B. Mitrovic

S. Rothstein

J. Vrbik

Computational Mathematics and Combinatorics [insert link here]

S. Anco

W. Craig

H. Fuks

S. Houghten

B. Ombuki

T. Wolf

Modelling, Simulation and Artificial Intelligence [insert link here]

J. Barchanski

D. Bruce

H. Gordon

B. Ross

S. Rothstein

T. Wolf

Statistical and Data Analysis [insert link here]

M-L. Huang

E. Sternin

Bioinformatics [insert link here]

H. Gordon

M. Richards

A.. Skandalis

R. Nadon

COMPUTATIONAL PHYSICS

Computational physicists use high-performance computers
to explore physical phenomena, from those involving the most fundamental
objects such as electrons and asteroids, to those important in applications
to areas such as materials science and medical technology. This new approach
to physics opens avenues to problems whose solution is otherwise impossible
and to gain physical insight needed to make those advances.

Most physical properties of gases, solids and liquids
are dictated by the behaviour of the electrons in the atoms that form the
building blocks of these systems. It is important that we study and understand
these electrons as accurately as possible, so that we can not only describe
the physical properties better, but also predict new phases and new types
of materials and fabricate solids and liquids with tailored problems. This
is a complex problem, residing at the heart of current materials science
technology, is being addressed by the research programs of BOSE, MITROVIC
and ROTHSTEIN.

VRBIK's reseach is focused on investigating the perturbed
Kepler equation to understand the resonances frequently found in our solar
system, and to investigate the solar system's stability and long-time behaviour.

------------------------end of computational physics team section------------------------------

------------------------- individual blurb (BOSE; include link to his
home page------

The research of BOSE and co-workers has the objective
of providing a fundamental understanding of the electronic structure of
non-crystalline materials and of transport properties of liquids and amorphous
metals, alloys, and semiconductors. In addition they do theoretical studies
of the spectroscopy of solids, collision-induced absorption, and vibrational
and magnetic properties of amorphous materials.

1.) PHASE TRNASITIONS AND SPIN/GUAGE GLASS PHASES IN QUASICRYSTALS.

Recently Monte Carlo simulations of localized spin models
on decagonal (Penrose) and octagonal quasiperiodic lattices were performed.
Many interesting results, including the possibility of a gauge spin glass
transition in two dimensional systems were obtained. This last result,
albeit controversial, has aroused much interest and definitely needs to
be explored further. Unfortunately this work is currently on hold, since
progress can be made only via extensive computational work requiring much
larger memory and computing speed than currently available to BOSE at Brock,
needed even to get the algorithms to the developmental stage.

2.) AB INITIO STUDIES OF ORDER-DISORDER TRANSITION AND SPINODAL DECOMPOSITION
IN RANDOM ALLOYS

In collaboration with European colleagues BOSE investigated
the ordering tendency in some random states of alloys using density functional
theory. Effective pair (EPI) and triplet interactions, calculated via the
coherent potential approximation generalized perturbation method, were
used to predict both primary and secondary transitions in the alloys.

A logical extension would be to use the EPIs calculated
via the density functional generalized perturbation method in an effective
Ising Model Monte Carlo simulation to describe the details of the transition,
including an accurate determination of the transition temperature and the
critical point exponents. Almost all numerical simulations of phase diagrams
in alloys are based on empirical parameters; studies based on electronic
structure have not as yet gone beyond the mean field approximation. Extension
along these lines is highly desirable, and improved computing resources
at Brock will enable them to develop the codes necessary to undertake this
research.

3.) AB INITIO STUDIES OF EXCHANGE INTERACTION AND OF MAGNETIC TRANSITIONS
IN PERIODIC AND GLASSY SOLIDS.

BOSE and co-workers recently studied the effect of topological
disorder on the magnetism of Fe and Co by using the linear muffin-tin orbitals
Green’s function method. This was the first ab initio study of the exchange
interactions in magnetic glasses. A future goal is to use the calculated
exchange interactions in a Heisenberg model Monte Carlo simulation to study
the magnetic phase transitions in Fe- and Co-based glasses. This calculation
would be time consuming as the exchange interactions are dependent on local
spin configurations, demanding new calculation of the exchange interaction
with every Monte Carlo move. However, this approach is valuable and novel,
as it is free from adjustable parameters and would provide the first application
of the knowledge of electronic structure in the study of magnetic transitions.

For more details on Prof. Bose's research, please consult
his home page. [insert link]

---------------------------------end of BOSE blurb--------------------------------

---------------------individual blurb (FUKS); include link to his home
page--

RESEARCH AREAS: discrete dynamical systems, mathematical
models of growth and diffusion phenomena in nonhomogeneous media (such
as bacterial growth in contaminated food products).

Traditional models of phenomena with complex spatial dynamics
are usually based on partial differential equations (PDE). They are quite
inadequate in situations where local stochasticity plays a prominent role,
and when the modeled system consists of many locally-interacting discrete
units, such as individuals in biological populations. To model such systems,
FUKS uses a novel class of tools, spatially extended dynamical systems
(SEDDS), including cellular automata and lattice gas automata. SEDDS often
exhibit simplicity expected from a good mathematical model, and at the
same time provide realism expected from a good simulation.

For more details on Prof. Fuks' research, please consult
his home page. [insert link]

---------------------end of FUKS blurb-------------------------------------------

-------------------individual blurb (MITROVIC); include link to his
home----

There are very strong physical arguments that in systems
with low superfluid density (e.g., high-T_c copper oxide superconductors
and the organic superconductors) the fluctuations in phase of the superconducting
order parameter play a significant role. Indeed, measurements of the ab-plane
magnetic field penetration depth gave unambiguous evidence for three-dimensional
XY critical scaling behavior below T_c. It is further argued that in the
case of poor screening, as indicated by a low conductivity, phase fluctuations
might influence the superconducting properties over a wide range of temperatures
below T_c. In this scenario a low-temperature order parameter would vary
linearly with T as found experimentally for the ab-plane magnetic field
penetration depth.

If, indeed, the phase fluctuations in the order parameter
are responsible for the observed linear temperature dependence of the ab-plane
superfluid density at low T, then recent experiments on c-axis electrodynamics
in YBCO are puzzling and have to be explained. Namely, it was found that
the c-axis penetration depth never has the linear temperature dependence
observed in ab-plane.

MITROVIC is examining the charging effects (i.e., the
effect of the Coulomb interaction between the charge density fluctuations)
on the temperature dependence of the c-axis penetration depth. Using a
quantum Monte Carlo algorithm developed for this purpose, he will compute
this quantity for the XY-model augmented to include the charging effects.
This calculation is fundamental as it will elucidate which processes determine
the electromagnetic response of high-T_c copper oxide superconductors.

For more details on his research, please consult Prof.
Mitrovic's home page. [insert link here]

-----------------------------------end of MITROVIC blurb-----------------------

----------------start of individual (ROTHSTEIN) blurb-------------------------

QUANTUM MONTE CARLO STUDIES OF PHYSICAL PROPERTIES

Traditionally there is a strong interest in energy-related
physical properties, such as the force constant and other spectroscopic
constants. It is a challenge to estimate physical properties other than
the energy, because the electron distribution obtained in quantum Monte
Carlo is not sufficiently accurate for these properties. In principle,
the exact electron distribution is required, the square of the unknown
exact wavefunction.

Although we do not have an analytic form for the exact
wavefunction, ROTHSTEIN and others have shown how to sample it by Monte
Carlo methods. Recently he developed a quantum Monte Carlo algorithm, practical
for properties represented by non-differential operators, where indeed
the "exact" electron distribution is sampled, with only a small error due
to the incorrect nodes of the guiding function.

His goal now is to derive within the framework of quantum
Monte Carlo a systemized methodology to estimate the non-trivial electrical
properties of molecules, such as high order polarizabilities and hyperpolarizabilities.
He is building on previous work in his laboratory, promising higher accuracy
for these properties, eventually employing methodologies to minimize the
nodal error. Polarizabilities are potentially fundamental in determining
the molecular geometry of products of chemical reactions, and they are
basic for understanding the electrical properties of advanced electronic
and photonic materials and devices, such as nanotubes.

Accurate low-order hyperpolarizabilities are available
for small systems (molecules up to 10 electrons) using analytical methods,
such as coupled cluster theory. It is quantum Monte Carlo which promises
an accurate and easy extension to higher-order properties for much larger
systems.

For more details on Prof. Rothstein's research, please
consult his home page. [insert link here]

-------------------end of individual link-------------------------------------

------------------start of individual (VRBIK) link------------------------

VRBIK works on new techniques of Celestial Mechanics (the
perturbed Kepler problem in particular), utilizing an older idea of regularizing
and linearizing the corresponding equation by means of Quaternion algebra.
This approach was extensively explored about 25 years ago, but only one
of its many advantages was clearly demonstrated when the new equation was
numerically integrated (little progress in solving it analytically). He
has developed an analytical (even though perturbative) approach to solving
this problem, which relies heavily on symbolic-algebra computations. One
can also extract a semi-analytic solution (Fourier series with numerical
coefficients); this approach being as computationally demanding as the
symbolic one (thousands of terms required for each quantity).

VRBIK has demonstrated the technique's feasibility and
usefulness in a variety of problems (lunar theory being the standard benchmark);
its advantages becoming apparent mainly for resonant forces (e.g. creation
of Kirkwood gaps), since it alleviates the issue of small divisors.

For more details on Prof. Vrbik's research, please consult
his home page [insert link here]

-----------------end of individual link-------------------------------------

----------------start of individual (WOLF) link------------------------

As symmetries of a Riemannian space are associated with
conservation laws of geodesic motion in this space, the investigation of
Killing vectors and Killing tensors is a central task in General Relativity

in classifying and understanding a spacetime.

Killing tensors are important too for polynomial first
integrals of motion described by Hamiltonian functions and for the separability
of

Hamilton-Jacobi equations in classical mechanics. An algorithm

developed by WOLF allows to write the integrability conditions of the

partial differential equations which determine Killing tensors as a

system of linear algebraic equations. The solution of these algebraic

systems was prevented so far by their size. New algorithms and

implementations to solve such linear algebraic systems shall make the
effective computations of Killing vectors and tensors possible.

For more details on Prof. Wolf's research, please consult
his web page. [insert link here; under construction]

---------------end of individual link------------------------------------------

---------------start of CM&C team section---------------------------------

COMPUTATIONAL MATHEMATICS AND COMBINATORICS

Fundamental physical phenomena have mathematical descriptions
in terms of equations. The objective of computational mathematics is to
study certain mathematical aspects of these equations and their solutions.
The results are intended to deepen the mathematical knowledge of such equations
and thereby carry across to physical understanding of phenomena that the
equations describe. In addition to applications in the physical sciences,
such as in material sciences, computational mathematics has been applied
to finance, biology, and system analysis.

A Webster-type definition of combinatorics is the study
of the arrangement of, operations on, and selection of discrete mathematical
elements belonging to finite sets or making up geometric configurations.
Combinatorical methods have been used to solve problems in telecommunications
and networking, cryptography, and in the design of computers.

Waves, light, gravity, and elementary particle interactions
are described mathematically in terms of wave equations and physical field
equations. The purpose of ANCO's, CRAIG's, and WOLF's research is to study
these kinds of equations, focusing in particular on symmetries, conservation
laws, and properties of their solutions.

When transmitting data, errors can occur due to noise.
To deal with this, error-correcting codes are used in many areas including
for satellite transmissions and to store data on compact discs. These codes
do not exist for all possible sets of parameters, so we may have to perform
a search to determine if that code exists, and if so, determine its structure.
This process is called a combinatorial search. The objective of HOUGHTEN
and OMBUKI research is to develop efficient algorithms and programs to
use in those searches.

------------------------end of link---------------------------------------------------------------

---------------start of individual (ANCO) link---------------------------------------------

COMPUTATION OF SYMMETRIES AND CONSERVATION LAWS

It is widely recognized that symmetries and conservation
laws are central to many aspects of analysis of differential equations.
ANCO's work involves both development of calculational techniques in this
area, and applications of the techniques to particular ordinary and partial
differential equations arising in applied mathematics. In general the calculation
of symmetries and conservation laws is a very non-trivial computational
problem, but in recent years the development of powerful computer algebra
programs has opened the door to tackling the necessary calculations.

For more details on Prof. Anco's research, please consult
his home page. [insert link here]

--------------end of individual link--------------------------------------------

--------------start of individual (CRAIG) link--------------------------------

APPLIED AND INDUSTRIAL MATHEMATICAL SCIENCES LABORATORY

CRAIG is the director of the Applied and Industrial Mathematical
Sciences (AIMS) Laboratory at McMaster University. One major undertaking
of the AIMS Lab is to establish a resource centre with infrastructure for
scientific calculations, providing a high performance computing environment,
state of the art scientific computational facilities, and a flexible laboratory
space for collaborative scientific work. The research activities
of the AIMS Lab will consist of mathematical modelling of processes of
interest to the physical and engineering sciences, most often in terms
of partial differential equations, and the subsequent mathematical analysis,
asymptotic analysis, and numerical simulations of solutions of these differential
equations.

For more details on Prof. Craig's research, please consult
his web page. [insert link here; @mcmaster]

---------------end of individual link---------------------------------------------

---------------start of individual (HOUGHTON) link--------------------------

COMBINATORIAL ALGORITHMS

In a combinatorial search HOUGHTON is attempting to determine
the existence of, or enumerate all, combinatorial objects with a given
set of parameters. A major difficulty in these searches arises from the
potentially huge search space and thus the use of computers is generally
a necessity if one wishes to complete such a search accurately and in a
reasonable amount of time. Due to the mathematical nature of these objects,
one must employ mathematics to understand the nature of the problem, and
thus sufficiently restrict the search space. The use of efficient algorithms
is not only of great importance, but also their design is of interest in
its own right.

For more details on Prof. Houghton's research, please
consult her home page. [insert link here]

---------------------end of individual link----------------------------------------------------------

-------------------start of individual (OMBUKI) link-------------------------------------------

PARALLEL COMBINATORIAL OPTIMIZATION TECHNIQUES In real-life, we encounter
combinatorial optimization problems in various situations. The potential
of providing innovative solutions for combinatorial optimization problems
by using modern meta-heuristics such as simulated annealing, tabu search
and genetic algorithms (GAs) has been examined. However, due to the complexity
due to conflicting constraints and huge search spaces by combinatorial
optimizations problem instances, there is need to consider high-performance
combinatorial optimization based on parallel and distributed environments.

The recent advancement of parallel processing allows one
to execute efficiently with low cost optimization algorithms on parallel
processing platforms (multi-processor systems). OMBUKI is investigating
parallel cooperative searching with meta-heuristics as an efficient approach
for solving complex combinatorial optimization problems.

For more details on Prof. Ombuki's research, please consult
her web page. [insert link here; it will be under construction]

----------------------end of individual's link------------------------------------------------

---------------------start of individual (WOLF) link--------------------------------------

1) COMPUTER ALGEBRA ALGORITHMS

A major part of WOLF's work is the development of computer
algebra algorithms which are applicable for the solution and study of systems
of ordinary and partial differential equations. The algorithms are designed
for solving over-determined systems of partial differential equations,
often very large, as arise in the calculation of symmetries, conservation
laws and variational principles, and linearization integrability for differential
equations. This work is helping to provide a powerful, general-purpose
tool useful in many areas of differential equations. The development of
algorithms goes hand in hand with work on specific applications,
which on one hand specify requirements for the algorithms and which are
also a good test bed for the algorithms and the programs developed. It
is planned to collaborate with S. ANCO on such problems.

2) OPTIMIZATION TECHNIQUES

An interesting new area is the use of stochastic optimization
techniques for solving combinatorial problems. Such techniques like genetic
algorithms and simulated-annealing are known to be useful to find approximate
solutions for very hard optimization problems but so far they have not
been used to solve combinatorial problems with singular discrete exact
solutions.

For more details on Prof. Wolf's research, please consult
his web page. [insert link here]

----------------------end of individual link here---------------------------

----------------------start of SM&AI team link here-----------------------

SIMULATION, MODELLING AND ARTIFICAL INTELLIGENCE

The term "simulation" applies narrowly to molecular dynamics,
molecular mechanics, and the Monte Carlo method, but more broadly also
to virtual environments and to animation. Through simulation one can visualize
concepts which are not obvious by using traditional methods; through modelling
one can experiment with physical models to arrive at desired outcomes quickly,
safely, and inexpensively. Related to this is the field of artificial intelligence,
which is concerned with developing software that accomplishes tasks normally
identified as requiring human intelligence, but are infeasible to accomplish
in reasonable time using human effort. These fields have the potential
to play a central role in the design of new materials and industrial processes
and in advancing our understanding of physical and biological processes.

Complex natural systems may be modelled as dynamical
systems. These models find many practical applications ranging from fluid
dynamics to biology and social sciences. The goal of FUK's and GRAIG's
research is the development of mathematical tools to help understand these
dynamics. This will allow better prediction and control methods for complex
systems, such as traffic flow, data networks, or growth and self-organization
of semi-conduction surfaces.

Mobile cooperative robotic teams can be considered to
be a kind of distributed computing system cooperating in execution of some
common tasks, reacting dynamically to changes in their environment. This
is a multidisciplinary topic involving wireless communication, mobile computing
and robotics. Using simulation methods, BARCHANSKI will discover team performance
in different circumstances, and moreover to find the robots' optimal communication
range, optimal number of robots for different tasks, and the influence
of robot learning algorithms on a team performance.

Antibodies are an important defense mechanism in the body.
GORDON's research will contribute important knowledge for the successful
engineering of novel antibodies for the purposes of medical diagnostics
or clinical applications. Her goal is to better understand how the three-dimensional
shape of antibodies contributes towards their ability to recognize, bind,
and thus target antigens for destruction. She is doing this by using computer
simulations to study how the inherent flexibility of the active site of
the antibody, called the hypervariable loop region, influences the antibody's
binding capabilities.

Photosynthesis provides the energy required for life on
earth via the conversion of absorbed sunlight into chemical energy, a process
called photochemistry. Although much of the mechanism of photosynthesis
is understood, some of the important details of light capture and energy
conversion are still missing. BRUCE is modelling one of the most important
regulatory mechanisms in photosynthesis, the one responsible for determining
the efficiency of photosynthetic energy conversion under natural conditions.

Multiprocessing and distributed computer systems are becoming
increasingly common. Unfortunately, software intended for such systems
is difficult to write, since multiprocessing systems are inherently complex
to control. ROSS's research focuses on the application of an aspect of
artificial intelligence technology, called genetic programming, to the
complex domain of concurrent computation.

NONE of the known techniques in Artificial Intelligence
or the whole Computer Science (like genetic learning, temporal difference
learning, machine learning of any form, pattern matching, tree searching,
parallel computing (even with say 10^6 computers), purpose build hardware,...)
can offer any breakthrough, the problem of Computer Go. This demands the
development of new methods and techniques. This is principally different
from other games, like chess. WOLF's research has the objective of designing
computer Go programs, which currently perform only at the beginners level,
which overcome the challenges presented by this game.

------------------------end of link----------------------------------------------------------------------------------------------start
of individual (Barchanski) link---------------------------------

SIMULATION STUDY OF MOBILE COOPERATIVE ROBOTIC TEAMS

BARCHANSKI is interested in mobile cooperative robotic
teams. Such teams can be considered a kind of distributed computing systems
cooperating in execution of some common tasks and reacting dynamically
to changes in their environment. This is a multidisciplinary topic involving
wireless communication, mobile computing and robotics.

While experimentation with real mobile robots would be his preference,
it is very expensive and time consumming. He is preceding it therefore
with extensive simulation study.

In his previous work BARCHANSKI proposed some algorithms
for inter-robot communication which he would like to study using simulation.
It would give him an idea about their performance in different circumstances.
He would like moreover to use simulation to find an optimal communication
range of robots' transmission when using single-hop or multi-hop routing,
optimal number of robots for different tasks and influence of robot learning
algorithms on a team performance.

For more details on Prof. Barchanski's research, please
consult his web page. [insert link here]

-------------------------end of individual link-----------------------------------

-------------------------start of individual (BRUCE) link----------------------

BRUCE's research which is most relevant BISC is a novel
implementation of genetic algorithms in the construction of a model for
water oxidation in photosynthesis. He has also modelled excited state
dynamics amongst the 36 chromophores of photosystem II whose positions
and orientations have recently been identified by X ray crystallography.
His laboratory is embarking on molecular dynamics modelling and further
excited state modelling of site directed mutants of photosystem II which
will affect specific chromophore position and orientation. Promising
“model mutants” will then be constructed and experimental data taken to
test against the molecular dynamic and excited state models. The
goal is to identify key chromophores in the regulation of photosynthetic
efficiency.

For more details on Prof. Bruce's research, please consult
his web page. [insert link here]

----------------------------end of individual link------------------------------------------------

----------------------------start of individual (GORDON)-------------------------------------

Antibodies constitute an important part of the defense
mechanism in the body. GORDON's research is contributing important knowledge
for the successful engineering of novel antibodies for the purposes of
medical diagnostics or clinical applications. Her goal is to better understand
how the three-dimensional shape of antibodies contributes towards their
ability to recognize, bind, and thus target foreign substances (antigens)
for destruction. She is doing this by using computer simulations to study
how the inherent flexibility of the active site of the antibody, called
the hypervariable loop region, influences the antibody's binding capabilities.
Most theoretical and molecular modelling studies of uncomplexed antibodies
have focussed on locating single low energy conformations of the six hypervariable
loops comprising the antigen recognition and binding site. However,
experimental results show that the antigen binding site is inherently flexible
and cast doubts that antibody selectivity and specificity is attributable
to a single low energy conformation of the hypervariable region.

While some computational work has been done to describe
local dynamical behaviour of hypervariable loops using molecular dynamics
(MD) simulations, the simulation lengths required for a complete description
of the equilibrium conformational distribution are not achievable.
GORDON's efforts are focussed on developing Monte Carlo sampling algorithms
that will more efficiently sample conformational space of peptide loops,
such that the conformation distribution of the hypervariable region will
be described. This will be necessary in order to correlate hypervariable
loop flexibility with antibody specificity and selectivity. The groundwork
is then prepared for computational experiments that will evaluate the effectiveness
of point or multiple residue 'virtual' mutations on antibody activity.
Thus the computational methods developed by her research will be of tremendous
use to the emerging Canadian biotechnology sector.

-------------------------------end of individual link-----------------------------------------------

-------------------------------start of individual link (ROSS)----------------------------------

EVOLUTIONARY COMPUTATION

ROSS' main area of interest in evolutionary computation
is formal language induction. He is interested in synthesizing formal grammatical
systems from examples of their intended behavior. For example, he has done
work in evolving concurrent systems implemented as process algebra. He
has also investigated the induction of stochastic regular expressions from
probabilistic examples of their behavior. Recently, he used this stochastic
regular expression language to evolve motifs for protein sequences, as
obtained from databases such as PROSITE.

In all this work, the machine learning paradigm ROSS is using is genetic

programming (GP). GP is an evolutionary computation technique in which
computer programs are evolved using an algorithm inspired by Darwinian
evolution. GP has a proven track record of being highly applicable to a
variety of different nontrivial problems.

ROSS' future research plans entail continuing the application of GP
towards language induction. He is interested in refining the evolution
of stochastic regular motifs. Preliminary results are very promising, and
he feels they can be improved further by enhancing the methods for optimization,
training, and motif language definition. I also intend to study the evolution
of a variety of concurrent systems, using specialized process algebra.

------------------------end of individual link---------------------------------

-------------------------start of individual (ROTHSTEIN)---------------

NOVEL APPROACH TO CHARACTERIZING PROTEIN STRUCTURES

Recent computational and experimental biological research
underline the importance of characterizing the ensemble of protein structures
corresponding to local minima in the potential energy surface. New Monte
Carlo algorithms are being developed to sample a wider phase space than
conventional methods with the objective of identifying non-global minimum
energy structures. It has been shown that protein models which accommodate
a multiple-conformation native state with substantial energy fluctuations
well below the unfolding transition temperature are consistent with experimental
measurements. This underscores the need for refined interpretations of
theoretical models taking into account native-state conformational diversity.

Recently ROTHSTEIN and co-workers published a novel pattern
recognition technique, “histogram filtering”, with which to optimize parameters
in wavefunctions for use in quantum Monte Carlo simulations. Its extension
to optimization problems involving Monte Carlo-generated data in computational
biology is immediate and obvious. ROTHSTEIN's research is exploiting histogram
filtering in conjunction with cluster analysis to a) characterize the low-energy
local minimum energy structures, and b) to arrive at a complete description
of the distribution of conformations for proteins, without having to take
recourse to a very large number of simulated annealing runs. While currently
completing this task “by hand” for a simple, yet highly-relevant model
protein, he will computer-automate the procedure to make feasible its implementation
on much more complicated structures, eventually with thousands of structural
parameters.

Notwithstanding advances in computer technology and geometry
optimization software, due to the complexity of the problem a theoretical
description of protein structure will inevitably be incomplete. The success
of this research promises a method to both uncover classes of low-energy
structures from a modest-sized data set and to characterize them. The resulting
gain in both computational efficiency and theoretical insight will extend
the frontier of computational biology problems amenable to investigation.

For more details on Prof. Rothstein's research, please
refer to his home page.[insert link here]

-------------------------end individual link--------------------------------------------------------------

-------------------------start individual (WOLF) link-------------------------------------------------

COMPUTER GO AND ARTIFICIAL INTELLIGENCE.

In computer Go, local search (tactics) and global thinking
(strategy) are interrelated; the game of Go does not represent a
single fight but involves a number of local fights going on in parallel,
usually interfering with each other. Thus the local tactical search
needs global strategic information, which in turn depends on the outcome
of local fights. This is why Go-programs still play only at a beginners
level, despite monumental effort spent by the artificial intelligence
community: genetic learning, temporal difference learning, all forms
of machine learning, pattern matching, tree searching, parallel computing,
purpose build hardware, etc. The situation is in distinct contrast to chess
where speed (purpose built hardware and massive parallelism) can compensate
human intelligence.

The problem of Computer Go therefore requires the development
of new methods and techniques. A promising new concept WOLF would like
to pursue is to characterize a game position as a dynamical system and
analyze this system of equations numerically. This is a novel approach,
requiring large computing power.

In Japan Go is played by the economic, financial and political
elite of the country as well as by about 7 million other people, and at
least as many in Korea and China. A computer Go program with the strength
of a professional player would not only represent a very large economic
asset, it would even have considerable political value.

For more details on Prof. Wolf's research, please refer
to his home page. [insert link here]

-------------------------end individual link--------------------------------------------------------------

--------------------------start B team link-------------------------------------------------------------------

BIOINFORMATICS

Biomedical research is an information-based discipline.
There is a major revolution in progress as novel experimental approaches
are yielding unprecedented amounts of data. The fields of medicine, biology,
and biotechnology are increasingly dependent on accessing this information.

Bioinformatics is an interdisciplinary field at the intersection
of life and information sciences which provides the tools and resources
for this endeavour. Development of innovative methodologies and practical
applications in this important field (such as gene discovery and genomics)
are objectives of the research programs of GORDON, RICHARDS, and SKANDALIS

------------------------end of link---------------------------------------------------------------

------------------------start of individual (NADON) link-----------------------------------------

NADON is currently the Director of Informatics at Imaging
Research Inc. He has lead a team of scientists and programmers who
have developed a software package designed for statistical analysis of
gene expressions arrays. This work required that novel methods be
developed to address various issues in array genomics data, notably very
small sample sizes, ubiquitous presence of outliers, and various sources
of systematic error.

SELECTED REFERENCES

a) Nadon, R., Shi, P., Skandalis, A., Woody, E., Hubschle, H., Susko,
E., Rghei, N., & Ramm, P. (2001). Statistical inference methods for
gene expression arrays. BIOS 2001 International Biomedical Optics Symposium.
San Jose, CA.

b) Nadon, R., Woody, E., Shi , P., Rghei, N., Hubschle,
H., Susko, E., & Ramm, P. (in press). Statistical inference in array
genomics. In Daniel Geschwind & Jeffrey Gregg Microarrays for the Neurosciences:
The Essential Guide. Cambridge, MA: MIT Press.

Click here for Imaging Research's homepage. [insert link
here]

------------------------end of link----------------------------------------------------------------------

------------------------start of individual (RICHARDS) link------------------------------

BIOINFORMATICS: PATTERNS OF GENE EXPRESSION

Correlating changes in patterns of gene expression that
lead to different types of behaviour is the essential goal of the field
of behavioural genetics, a field that is evolving from simple genetic mapping
of behavioural mutants to sophisticated investigations of the influence
of genes on behaviour, and of behaviour on genes.

Learning how genes influence the expression of altruism
in social sweat bees is the fundamental goal of RICHARD's research.
Halictine sweat bees are the most socially variable of any animal group,
including species ranging from completely solitary to strongly eusocial,
and even including facultatively social species. In all social species,
newly emerged females are ‘totipotent’, that is, capable of acting as either
queens or workers. Which caste they eventually join is strongly influenced
by larval nutrition and genetic relationships among nestmates. By
observing female sweat bees from larval to early adult stages, she can
investigate how changes in gene expression influence and are influenced
by the development of caste-specific behaviour. Particularly crucial
is to investigate how different patterns of gene expression are associated
with the expression of altruism, behaviour that typically is expressed
only by workers. Recent technological advances in honeybee genomics
have created molecular tools that can be used for sweat bee studies.
At the DNA sequence level, strong similarities between honeybees and sweat
bees allow her to use tools such as honeybee DNA chips to analyze gene
expression patterns in sweat bees using microarray analysis. This
means that elucidating genetic mechanisms determining behaviour, especially
those underlying the specialized caste behaviours of queens and workers,
is an achievable and worthwhile goal.

A second avenue of research is the use of DNA sequences
for analyzing phylogenetic relationships among closely related sweat bee
species, and in fact, for identification of cryptic species and subspecies,
the existence of which greatly complicates the interpretation of behavioural
data unless the species are confidently delineated. DNA sequence
data are also being generated for molecular evolutionary studies of evolutionary
rates (in terms of DNA and amino acid substitution rates) and mutational
rates of haplodiploid vs diploid insects, a new line of research recently
opened up in her research group.

RICHARDS is dealing with the mountains of microarray and
DNA sequence data she is currently generating and studying. She is
focusing on the development and expression of caste-specific behaviour
in social sweat bees, using microarray analyses for both intra- and inter-specific
comparisons of gene expression patterns. In honeybees and sweat bees,
differences in larval nutrition appear to trigger a developmental switch
that leads to the expression of the caste-specific behavioural repertoires.
The timing of this switch and the nature of the genes that are turned off
or on in queen- and worker-destined females will be investigated using
microarray analysis of queen and worker-destined bees from early larvae
to 3 days post-eclosion. Once candidate loci have been identified,
the study will be enlarged to compare pairs of solitary vs. social sibling
species to determine whether particular loci may be responsible for facultative
variation in sociality.

------------------------end of individual link---------------------------------------------------------------

------------------------start of S team (Huang) link----------------------------------------------------------

STATISTICAL ANALYSIS

Nonparametric regression and ANOVA are important
research directions in last 20 years with many applications. HUANG's research
is developing new kinds of quantile, regression estimation, prediction
and ANOVA methods. The mathematical properties of these estimators, predictors
and test statistics are being studied: consistency, rate of convergence,
efficiencies. She is building stochastic models based on these methods
and applying them to economics, sciences, quality control, telecommunication
network and biostatistics.

Studies of truncated and censored data
have important applications in biostatistics, health studies, industrial
engineering and other fields. Her research will develop new nonparametric,
Bayesian and likelihood methods appropriate for such data. The work also
links to combinatorial occupancy models and theory.

The above research utilizes computer
simulations, bootstrap re-sampling in large data bases, and spectral analysis
of time series.

For more details on Prof. Huang's research, please consult
her homepage. [insert link here]

------------------------end of individual link-------------------------------------------------------