 This summer,
twenty-five undergraduate student researchers from the University and around the
country served ten-week internship appointments at the Supercomputing Institute.
The students were selected from a pool of over one hundred applicants to participate
in programs in biophysical computing and computational dynamics as well as scientific
computing and graphics. The students worked closely with faculty advisors on projects
ranging from motion of a chain of rigid bodies to the modeling of a stent.
The Summer Internship Program is sponsored by the Supercomputing Institute and the
National Science Foundation's Research Experiences for Undergraduates Program, which
is in its eighth year. The program promotes undergraduate involvement in ongoing
and new research in many fields and provides students with an opportunity to work
full-time on challenging and computationally intensive problems in an academic research
environment.
During the course of the summer, the students participated in Institute sponsored
tutorials specific to high-performance computing and in individual laboratory tours
led by faculty members. To conclude the summer, the students presented talks open
to the entire research community. These talks allowed them to share their work with
other researchers and to gain experience making scientific presentations. The program
allowed the students to perform research in close collaboration with faculty investigators
and their research groups and to discuss research with faculty members, post-doctoral
associates, graduate students, and other interns with similar interests.
Project Descriptions
Stefan Debbert, a Chemistry major at the University of Minnesota, worked with
Professor Christopher Cramer of the Chemistry Department on computational chemistry
approaches to understanding what influences effectiveness of antibodies on tumor
cells. Since tumor cells are consistently more acidic than normal cells under certain
physiological conditions, an antibiotic that is more reactive in the protonated form
(one extra proton) than the unprotonated form may be able to preferentially attack
tumor cells as opposed to healthy cells. Stefan looked at differences in pyridyne
and benzyne electronic structures brought on by nitrogen's lone pair to observe effects
and help determine the effectiveness of some of the antibiotics.
Derek Dolney worked with Professor Cramer and Professor Donald Truhlar of
the Chemistry Department on calculating solvation free energies with the Conductor-Like
Screening Model (COSMO), a method used to compute electrostatic portions of solvation
free energies. Derek carried out the initial steps of developing the SM5CR solvation
model (an implementation of the COSMO method) and incorporating it into the AMSOL
software package. Performance of the model was evaluated by testing its accuracy
for prediction of solvation free energies. The computation time required to complete
such calculations with various choices of parameters was evaluated, and an optimum
compromise of accuracy and cost was devised. Derek is majoring in Chemistry and Physics
and minoring in Spanish at the University of Minnesota.
Mala Radhakrishnan is majoring in Chemistry and Physics and minoring in Philosophy
at Harvard University. She worked with Professor Donald Truhlar of the Chemistry
Department on calculating reaction rates. The first step toward a theoretical reaction
rate calculation is optimizing geometries and determining electronic energies of
the species involved. Mala first optimized the geometry of the reactants, products,
and transition states of methanol, ethanol, and 2-propanol reactions using semi-empirical
methods incorporated into the
AMSOL computer package developed at the University
of Minnesota. Using the methanol reaction as a prototype for more complex reactions,
she analyzed accuracy, efficiency, and expense of other methods. A hybrid of the
Hartree-Fock and density-functional theories was found to give sufficient flexibility
to adjust calculated energies to agree with experiment. This combination was used
to calculate the reaction rate of hydrogen with methanol in aqueous solution, and
the results agree with experiment within experimental error. Nevertheless, future
work is planned to explore the effect of nonequilibrium solvation on the tunneling
contributions.
Jocelyn Rodgers, a Chemistry and Physics major at Harvard University, also
worked with Professor Truhlar. Jocelyn calculated reaction rate constants for the
reaction of hydroxyl (OH) radicals with alkenes and aromatic hydrocarbons. The OH
radical is the single most important radical in combustion reactions. Jocelyn's project
used the GAUSSIAN94 electronic structure package and the POLYRATE dynamics code for
reaction of OH radicals with small unsaturated hydrocarbons. The applicability of
this scheme for treating the reaction of OH with larger unsaturated organic compounds
was a long-term goal. The electronic structure calculation involved as low an order
of theory and as small of a basis set as possible to keep the expense of computing
time down. This scheme can then be applied to larger hydrocarbons without being prohibitively
expensive.
David Dreytser worked with Professor David Thomas of the Biochemistry Department
on modifying the three-dimensional, atomic model of phospholamban, a fifty-two amino
acid integral membrane protein of the sarcoplasmic reticulum. David wrote an algorithm
to generate the relative orientation of helices in a bundle with desired geometries.
This process required a reference helix chosen by finding the helical coordinate
system of the first helix. Orientations of remaining helices were found by rotating
every helix in the pentamer by identical angles. This process was repeated and the
final structures were used to deduce a structure of phospholambam. David is majoring
in Chemical Engineering and Biochemistry and minoring in Management at the University
of Minnesota.
Ethan Bernard, a Biochemistry and Biophysics major at Oregon State University,
worked with Professor David Levitt of the Physiology Department. A general method
for simulating the motion of a chain of rigid bodies using a recursion relation had
already been detailed. A recursion relation allows motion of an entire chain to be
defined by motion of the first body and motion in the links connecting successive
bodies. This method is different from a direct approach that simultaneously describes
the chain with a system of equations and solves the equations to determine motion.
Ethan wrote a program that simulated motion of the chain of rigid bodies using the
direct method and tested this against a program using the recursion relation.
Seth Gammon, a Biophysics major at the University of Illinois at Urbana Champaign,
also worked with Professor Levitt. Seth investigated characteristic change in chemical
shift of a residue in a particular type of secondary structure. He used an automated
approach to compare relationships between secondary structure and chemical shift.
More specifically, beta turns, gamma turns, beta hairpin turns, alpha helixes, and
beta sheets were looked at. Files containing Nuclear Magnetic Resonance (NMR) data
were collected and narrowed down by specific criteria. If there was a choice between
NMR ensembles or NMR averaged structures, the averaged structure was chosen. Seth
wrote a C++ program to match up chemical shift data with appropriate structural information.
Differences between standard chemical shift and reported chemical shift were determined
and the data was statistically analyzed.
Eden Paster, a Biochemistry and Biology major at Rice University, worked with
Professor William Gleason of the Laboratory Medicine and Pathology Department. Eden
used a known three-dimensional structure of horse cytochrome c, determined by X-ray
crystallography, to model for mouse cytochrome c. The amino acid sequences only differed
by six amino acids, which allowed Eden to mutate the horse sequence into the mouse
sequence. It was then possible to obtain a three-dimensional structure for mouse
cytochrome c by submitting the amino acid sequence to a modeling program. The structure
was subjected to energy minimization and compared to structures obtained by different
procedures. Eden then modeled the mouse cytochrome c interacting with mini-antibodies.
A three-dimensional model of the antibodies was constructed, and quantitative data
for a number of cytochrome c and miniantibody complexes was obtained.
Nhi Tran also worked with Professor Gleason. Nhi studied the three-dimensional
expansion of a Palmaz-Schatz stent, a mechanical device used during balloon angioplasty
(see www2.msi.umn.edu/Bulletin/Vol.14-No.3/july98.html). The system was modeled
by finite element analysis methods as implemented in the commercial MARC/MENTAT package.
Specific investigation dealt with the possibility of coating stents with biomaterials
suitable for local drug delivery of therapeutic agents from the stent. Flexibility
and resilience are required of stents to endure both deployment procedure and dynamic
forces of cardiac contractions. These performance requirements translate into longitudinally
flexible, corrosion-resistant, thromboresistant stents that need to be radially noncompliant,
of high-expansion ratio, and in complete contact with the vessel wall. Nhi is majoring
in Biomedical Engineering and minoring in Computer Science at Duke University.
Eric Hemmesch and Timothy McMurry worked with Professor Edward Egelman
of the Cell Biology and Neuroanatomy Department on the installation and configuration
of linux and supporting programs on a dual-processor computer. Their research began
by looking at interactions between biological proteins and muscle fibers (actin and
myosin). Electron micrographs allowed Eric and Timothy to visualize their conformations
in a two-dimensional manner. Although these two-dimensional pictures offer great
insight, they afford little information as to what occurs on the molecular level.
Because of this, there is a need for a program to transform these images into three-dimensional
structures that can be manipulated for a greater understanding of the atomic interactions.
Eric and Timothy used Fourier-Bessel analysis in an attempt to recover three-dimensional
information regarding the structure of the proteins. Eric is majoring in Chemical
Engineering and Chemistry and minoring in Management at the University of Minnesota,
and Timothy is majoring in Mathematics and Physics at Carleton College.
Gregory Wilde, a Biomedical Engineering major at Tulane University, worked
with Professor David Thomas of the Biochemistry Department. Gregory helped determine
the feasibility and applicability of molecular dynamics simulations for predicting
electron paramagnetic resonance (EPR) spectra of spin-labeled myosin. In order to
generate an accurate molecular dynamics trajectory, the project first found the most
stable conformation and potential function of a spin-labeled myosin. This required
a DISCOVERscript to minimize steric hindrance. A quick yet accurate model was decided
upon. Once the structure was determined and recognized, a simulation was run to output
orientation history of the spin label. If the system reached equilibrium, an order
parameter was calculated to produce an EPR spectrum, and results were compared.
Michael Enz, a Physics and Computer Science major at the University of Minnesota,
worked with Professor J. Woods Halley from the School of Physics and Astronomy on
experiments involving low-temperature beams of atomic helium. Michael reproduced
computer simulations of an experiment that applies current pulses to a resistive
element covered by an adsorbed helium film creating a helium beam of evaporated atoms
detected a few centimeters away by a superconducting bolometer. Michael's simulation
assumed the evaporated atoms obey a classical, Boltzmann distribution in velocity.
Data with multiple peaks in the detected signal from high-power source pulses with
thicker helium films were successfully described by a quantum evaporation theory
based on unusual dispersion curves for quasi-particles in liquid helium.
Seth Van Oort, a Computer Science major at the University of Minnesota, worked
with Professor David Lilja of the Electrical and Computer Engineering Department.
Professor Lilja's group has been developing a program that takes data from a running
program and displays it in an easily understood format. Seth worked on a segment
of the project that would display data on the screen. A tree is the natural way to
display methods. As the mouse is moved over each node, the data for that node is
displayed. One of the major tasks involved in displaying the data was writing a function
to search through a tree for nodes fitting certain parameters.
The source of some of the highest energy cosmic rays in our galaxy is still in question.
The prevalent theory was that these high-energy particles are accelerated in the
shock wave of a Supernova Remnant. Johan Hoff, majoring in Aerospace Engineering
at the University of Minnesota, worked with Professor Thomas Jones of the Astronomy
Department on a computer simulation of the progressive state of shock wave development
and a numerical calculation of cosmic ray momentum distribution that could be generated
by the modeled conditions. The simulation provided a calculable solution that accounted
for observed and new theoretical results of high-energy cosmic ray studies. The simulation
was performed on a hybrid, fluid-Boltzman-equation shock wave simulation code modified
to take the effects of cosmic ray pressure and momentum in development of the shock
wave into account.
Michael Greminger, a Mechanical Engineering major at the University of Minnesota,
worked with Professor Charles C.S. Song of the Civil Engineering Department. Michael's
project converted an existing
FORTRANprogram into a parallel program that could
run on a multiprocessor supercomputer or a network of workstations. He hoped conversion
would increase speed allowing larger problems to be solved. Michael first tried automatic
parallelization tools native to the supercomputers. He then tried parallelizing the
program using Message Passing Interface, a library of functions that can be used
to create parallel programs. Both methods were analyzed and compared.
Jeffrey Sommers worked with Professor Alon McCormick of the Chemical Engineering
and Materials Science Department on three projects. The first dealt with the distance
matrix, a mathematical entity that has proven useful in many fields. This work developed
a program that used an efficient algorithm to compute the distance matrix for a given
configuration of a certain number of atoms. The second project dealt with energies
of various cyclic structures. The final project worked with diffusion simulations.
Jeffrey began by writing a program that simulates diffusion in which the atoms can
pass each other. The program was then modified so atoms could not pass each other.
Jeffrey is majoring in Chemical Engineering and minoring in Chemistry at the University
of Minnesota.
Steven Miller, a Chemical Engineering major at the University of Notre Dame,
worked with Professor George Wilcox of the Pharmacology Department. Steven's project
dealt with resolution of structures on sub-micron scales by confocal microscopy.
This method is of value since a thick specimen may be optically sliced without damaging
its structure, and images may be combined to render an object in three dimensions.
The presence of out of focus light, noise, and other aberrations limit the usefulness
of images for quantitative analysis. Mathematical deconvolution restores the image
to a better representation of the true object. An object of known morphology containing
minute three-dimensional structures can test reliability of restoration and resolution
of the microscope. A glass micropipette containing fluorescent dye was imaged at
the tip where separation between compartments approaches the limit of resolution.
By comparing restored images to structures established by electron micrograph, suitability
of the method to quantitative analysis was determined.
Justin Sytsma also worked with Professor Wilcox. Justin is a Neuroscience
and Computer Science major and Philosophy minor at the University of Minnesota. He
simulated neurons using several mathematical models for comparison. One model was
used for peripheral unmyelinated nerve fibers of the rabbit sciatic nerve. This model
looked at a single point along the axon. Action potential was generated using a sodium
current to produce initial depolarization. A leak current was used for repolarization.
Once the best model was determined and code for the simulation was optimized, Justin
used the model to look at effects on action potential propagation of the application
of drugs to these fibers.
Rashid Zia worked with Professor David Yuen of the Geology and Geophysics
Department and more than a half dozen researchers to transform portions of their
models and simulations into a more communicative format. In the process, he became
very familiar with several visualization programs. Aside from pure visualization,
Rashid gained familiarity with both individual models and simulations of several
researchers and the general styles of computer simulations. He has also become acquainted
to general geothermal convection simulations, molecular dynamics simulations, and
well-mixed data analysis models. Rashid is majoring in Electrical Engineering and
American Literature at Brown University.
Andrew Shallue is majoring in Mathematics and minoring in Computer Science
at Gustavus Adolphus College. He also worked with Professor Yuen. Andrew's project
dealt with fractals and fractal theory. The most useful property of a fractal is
its fractal dimension, which measures the ruggedness of a natural formation or graph.
Andrew quantified the mixing process in mantle convection using fractal dimension
and compared dimension over time with different mixing processes. Earlier work had
analyzed mantle convection fields comparing differences in newtonian and non-newtonian
rheologies over time. However, results were suspected to be inaccurate since fields
are used more for visualization than analysis. With the development of the line method
of modeling mantle convection, it became possible to more accurately quantify mantle
mixing.
Christopher Messer, a Mathematics major at the University of Minnesota, worked
with Professor Dennis Hejhal of the Mathematics Department. Professor Hejhal's research
has illustrated chaotic phenomena in solutions to Schrödinger's equation for
a quantum-mechanical particle of specific energy. Christopher studied the occurrence
of this chaos within quantum-mechanical systems. He observed the solutions to Schrödinger's
equation for a quantum-mechanical particle of specific energy that became increasingly
chaotic as energy increased. Understanding the sources of this chaotic behavior led
to a study on why such randomness can be present even in highly regular two-dimensional
regions. Simplified cases with high symmetry in classical geometry were studied to
aid identification of the factors that contributed to randomness.
Robert Roos, a Computer Science major at Stanford University, worked with
Professor George Wilcox of the Pharmacology Department. Robert's project investigated
feasible ways of counting neurons. Many questions could be answered if a fast, reliable
method of counting certain cells was available. In some cases, these questions involved
the correlation between numbers of cerebellar Purkinje neurons and motor coordination
of mice at varying ages and pathological states. While counting is in some ways the
most basic type of data-gathering possible, it is extremely time-consuming. Robert's
project aimed to automate the process, requiring that one of the central problems
of computer vision-image recognition-be solved, to a high degree of accuracy, for
the specific types of cells to be counted.
Heidi Basler is majoring in Computer Science and Mathematics and minoring
in Physics at the University of Nebraska. She worked with Professor Leonard Banaszak
of the Biochemistry Department in designing World Wide Web pages. Heidi learned to
program in CHIMEsoftware, a plug-in that enables three-dimensional molecular structures
of a protein to be placed on a World Wide Web page. Unlike other programs, CHIME
displays live molecules that can be rotated and reformatted by users. CHIME allows
text explaining proteins to be added with buttons to manipulate proteins and highlight
regions, residues, and individual atoms.
Thomas Grys, a Biochemistry major at Gustavus Adolphus College, worked with
Professor Douglas Ohlendorf of the Biochemistry Department on creating two sets of
World Wide Web pages. One set provided descriptions of projects underway in Professor
Ohlendorf's research group. These pages included active three-dimensional displays
of macromolecules and links to other resources used by the group. This information
is for candidates interested in studying at the University of Minnesota and the general
public. The other set of pages was created for the 1999 International Conference
for Biological Inorganic Chemistry being organized by faculty at the University of
Minnesota. These pages allowed for online registration and submission of abstracts.
A mechanism was set up to allow interested parties to search the archived abstracts
for information about the projects and authors.
| 1998 Summer Tutorials, Laboratory Tours, and Intern Seminars |
Introduction to the Supercomputing Institute
Introduction to Scientific Visualization
Introduction to Perl
Introduction to the IBM SP Supercomputer
Introduction to LoadLeveler and Batch Job Submission on the IBM SP Supercomputer
Math/Numerical Libraries for the IBM SP Supercomputer
Molecular Visualization Tools
Single Processor Tuning for the IBM SP Supercomputer
Data-Parallel Code Development on the IBM SP Supercomputer |
Introduction to InsightII/Discover
Introduction to Message Passing Interface
Point to Point Communication with Message Passing Interface
Collective Communications with Message Passing Interface
Advanced Message Passing Interface
Introduction to the SGI Origin 2000 Supercomputer
Math/Numerical Libraries on the SGI Origin 2000 Supercomputer
Data-Parallel Code Development on the SGI Origin 2000 Supercomputer
Introduction to Java |
Introduction to Shell Programming
Advanced Unix Features
Introduction to Parallel Programming
Astronomy laboratory tour
Biochemistry laboratory tour
Cell Biology and Neuroanatomy laboratory tour
Chemistry laboratory tour
Pharmacology laboratory tour
Physiology laboratory tour
Summer Intern Seminars |
| |
|
Future Program Information
The Supercomputing Institute is pleased to announce our summer 1999 internship
program, open to both University of Minnesota and non-University of Minnesota undergraduate
students. This program provides undergraduate students an opportunity for a challenging
and enriching educational experience.
This internship is intended for undergraduate students interested in pursuing graduate
or professional education and research in scientific computing and graphics. Students
work with faculty on a wide variety of projects. Faculty from various disciplines
have contributed projects and are responsible for directing the students in their
daily work. Summer appointments will run from June 14 to August 20, 1999.
The Supercomputing Institute will also be providing Undergraduate Internships in
Winter and Spring 1999. These internships will be available to University of Minnesota
undergraduate students. Research projects will be available in a variety of disciplines.
The winter program will run from January 4 through March 12, 1999. The spring program
runs from March 29 to June 4, 1999.
Please check the Supercomputing Institute World Wide Web pages for more information.
|
|
|
