Volume 18, Number 1
Winter 2002 Research Bulletin of the Supercomputing Institute

2001 Undergraduate Summer Interns

Last summer, twenty-one undergraduate student researchers served ten-week internship appointments at the Supercomputing Institute. The students were selected from seventy-three applicants nationwide. The students performed research in close collaboration with faculty investigators and their research groups

Currently in its eleventh year, the Supercomputing Institute Summer Intern-ship Program promotes undergraduate involvement in ongoing and new research. Areas of study include scientific computing, digital technology, visualization in the physical, medical, and social sciences, and engineering, as well as software development for scientific computing and graphics in these areas. The program provides an opportunity for a challenging and enriching educational experience for undergraduate students interested in pursuing graduate or professional education and research in scientific computing and/or graphics.

During the summer, interns participated in Institute-sponsored tutorials specific to high-performance computing. Towards the end of the summer, the interns presented talks open to the entire research community that allowed them to share their work and gain experience making scientific presentations.

Project Descriptions

Left to right: Dawn Schafer, Clair Chisolm, and Victoria Bedell enjoy the program's opening activities.

Anthony Anderson, majoring in Chemical Engineering and Mathematics at the University of Minnesota, has been working under the supervision of Dr. Vittorio Cristini and Professor Christopher Macosko (Department of Chemical Engineering and Materials Science), and Professor John Lowengrub (School of Mathematics). Anthony developed adaptive three-dimensional numerical simulations of micro-structured materials, with applications to multiphase polymeric flow as well as to biosystems such as tumors and their tissue environment. The main focus of Anthony's work was on the development of Fortran routines for volume mesh reconstruction, to be employed on unstructured meshes of tetrahedra. The goal of the adaptive strategy was to achieve a mesh distribution that resolves the relevant length scales in the specific problem with prescribed accuracy. This involves three distinct steps: 1) dynamic equilibration of the marker node positions based on an analogy to a physical system of springs; 2) refinement/coarsening via node addition and subtraction; 3) mesh reconnection via edge swapping to achieve nearly equilateral discretization. Anthony made significant progress on all of these steps, and completed routines for mesh equilibration and refinement which are now being tested.

Majoring in Computer Science at the University of Minnesota, Abdul Bahar worked with Professor Douglas Ohlendorf (Department of Biochemistry, Molecular Biology and Biophysics). During the summer, Abdul wrote a Perl script (for a UNIX system) to automate map generation procedures. After giving it only one protein database file as input, his program generated maps, mask, and electron density files. This helped his research group save time by not importing the output of other procedures manually. He wrote procedures to generate identity matrices, multiply matrices, inverting 3 ´ 3 matrices, and using Cramers' rule to solve linear equations. Abdul also assisted in furthering the implementation of the Gaussian Elimination to solve linear equations.

Jedediah Deitrick, Peter Holm, AJ Klein Osowksi, and Abdul Bahar await the beginning of a tutorial.

A Biochemistry major from Bethel College, Victoria Bedell worked with Professor Steven Kass (Department of Chemistry). Victoria worked on research that involved carrying out high-level calculations on zwitterionic compounds (i.e., molecules which have a positive and negative charge site). Since all twenty essential amino acids exist in their zwitterionic form in water over a wide pH range, learning about these species is essential to gain a better understanding of protein structure and function. More specifically, Victoria assisted in carrying out geometry optimizations and vibrational frequency analyses on a series of methyl pyridinium anions, radicals and conjugate acids.

Sarah Betterman, majoring in Chemical Engineering at the University of Minnesota, worked with Professor David Grant (Department of Pharmaceutics). Sarah's research involved working with the software package Cerius2 to determine crystal structures from powder diffraction data using a compound called warfarin sodium. This drug has been used as rat poison and an anticoagulant for many years, yet its crystal structure has never been solved. This research group is trying to predict the crystal structure in order to explain the compound's solid-state properties and chiral discrimination in drugs and other molecules.

Nenad Bjelogrlic, majoring in Aerospace Engineering at the University of Minnesota, worked with Professor Fernando Porté-Agel (St. Anthony Falls Laboratory and Department of Civil Engineering). Nenad worked on a project that used an existing large-eddy simulation (LES) code to study the three-dimensional unsteady turbulent transport of heat, water vapor and pollutants in the atmospheric boundary layer (lowest part of the atmosphere, in direct interaction with the land surface). AMIRA and Iris Explorer were incorporated in the LES code, and the visualization process was automated. These tools were then used to visualize the dynamics of the atmospheric flow. This information was used by Portˇ-Agel's research group to identify and study the effects of different factors, such as, land surface heterogeneity and solar radiation on the transport rates. Nenad continues to work with the group on a web-based display of the visualizations (see: les.safl.org).

Adam Butensky-Bartlett, majoring in Integrated Science/Biophysics at Northwestern University worked with Professor David Thomas (Department of Biochemistry, Molecular Biology and Biophysics). Adam's major research activity focused on simulation of Electron Paramagnetic Resonance (EPR) spectra from molecular dynamics trajectories, with a goal of accurately simulating an EPR spectrum. EPR is a technique that studies the structural change of proteins (specifically the muscle proteins myosin, actin and the calcium pump) via the attachment of a spin-label to natural or artificially mutated cysteine residues. The result of this experiment was a graph (in frequency space) demonstrating the resonant frequencies of the spin-label, which vary between different cysteine attachment sites and are characteristic of the probe's local environment.

Mira Chaurushiyan, a Biology major at Carleton College, worked with Professor William Gleason (Department of Laboratory Medicine and Pathology). Mira worked in two major areas of research: the parallelization of protein database search algorithms, and the use of NAMD/VMD, an interactive molecular dynamics simulation package. NAMD is a parallel, object-oriented molecular dynamics code designed for high performance simulation of large, biomolecular systems, and was developed by the Theoretical Biophysics Group at the Beckman Institute at the University of Illinois. Dr. Gleason's research group uses Lutefisk, SEAQUEST, and other software to generate lists of potential peptide fragment sequences from experimental mass spectral data. Parallel versions of CIDentify (a protein database search program) then is used to search for these possible fragments in a database of known proteins to determine if they are present in any previously sequenced proteins. Using Message Passing Interface, parallel versions have been run on multiple processors of the SGI Origin and the 32-processor Netfinity Linux cluster at the Supercomputer Institute. Benchmark results indicate that a Linux cluster (on the order of eight processors) could be used, both for the software package that does this database search, and for the control of a mass spectrometer. If the potential sequences for a particular peptide fragment were found to not match any sequence in the protein database, and if this determination were made quickly enough, the machine searching the database could also control and modify the experiment automatically to obtain more information about the fragment.

Eric Johnson and Mira Chaurushiyan discuss the program during a luncheon break.

University of Tuscaloosa Chemistry major Claire Chisolm worked with Professor David Thomas (Department of Biochemistry, Molecular Biology and Biophysics). Claire worked with Dr. Thomas's research group on Electron Paramagnetic Resonance (EPR) simulation from molecular dynamics trajectories, and was primarily involved with creating models of a spin-label that the lab regularly attaches to samples in order to run EPR spectroscopy. She used the InsightII/Discover software to build a molecule and optimize its structure. In addition, Claire worked on the actual EPR simulation program. The goal was to be able to create a simulated EPR spectrum from molecular dynamics trajectory data, and have it correlate to the experimental data. Her portion of this project was to code a Fast Fourier Transform routine in C in order to transform the magnetization trajectories (calculated from molecular dynamics data) into EPR spectra. Claire and Dr. Thomas' research group found that it is possible to directly compare experimental and simulated EPR results, increasing clarity and understanding of the experimental results

Interns Jedediah Deitrick (majoring in Computer Engineering at Kansas State University) and Peter Holm (majoring in Computer Engineering and Mathematics at Iowa State University) worked with Professor David Lilja (Department of Electrical and Computer Engineering). Both students spent their internship rewriting computer performance benchmark programs from the SPEC2000 benchmark suite. They used profiling tools to determine the function call patterns of the programs, identified loops that could be parallized (if any), and rewrote the code using the machine instructions for the superthreaded processor (a new computer architecture under development in Professor Lilja's group). Jedediah and Peter tested their parallel code using the simulator for the superthreaded processor. Once the parallel code was running correctly, they examined how much faster their parallel code ran (as compared to the original code) by looking at the instructions executed per cycle and other performance metrics. The work that Jedediah and Peter conducted during their internship will be incorporated into publications produced over the next year by Professor Lilja's research group.

Zachary Garbow, majoring in Computer Science from the University of Minnesota, carried out data mining research with Professor David Yuen (Department of Geology and Geophysics). Data mining is a research approach used to extract relevant data from particular locations in much larger data sets in both 2-D and 3-D. Zachary used an assortment of computer languages to develop graphical user interfaces in the form of Java applets, thus allowing users to interact in a client/server paradigm. By implementing web-based applications for data mining in geophysical data they have shown how data mining can effectively parse massive data sets to display results that are easy to understand and analyze. Their methods provide a new way of interrogating various forms of geophysical data and visualize the results in a fast and efficient manner.

Michael Garrels from Case Western Reserve University enjoyed the Wednesday luncheons.

Michael Garrels, majoring in Materials Science and Engineering at Case Western Reserve University, worked with Professor Woods Halley (Department of Physics). Michael studied ion conduction in amorphous polymeric electrolytes by conducting a classical simulation of the resonant frequencies of the triflouromethanesulfonate (triflate) ion. Modeling the atoms as balls and the bonds as springs, he used a harmonic potential in internal coordinates and expanded and translated it into Euclidean displacement coordinates. Michael wrote code that randomly varied the values of spring constants and compared the spectrum produced to the experimentally observed spectrum. Michael then chose the parameters, accordingly.

Jerome Hauser of Princeton University.

Majoring in Computer Science at Princeton University, Jerome Hauser worked with Professor Shir Ramaswamy (Department of Wood and Paper Science). Jerome worked on streamlining the image analysis program that they have developed for visualizing and characterizing the three-dimensional (3-D) structure of porous media, and also helped develop user documentation for the image analysis software. Using the new systematic procedure and user documentation, he analyzed the 3-D structure of tissue and towel samples and then compared the results with other samples of different structure. He also worked on modifying another program to characterize the 3-D structure of porous materials using novel parameters, namely node density, coordination number and bond length distribution. Jerome also helped analyze the internal structure of a polymeric carrier fabric used in paper manufacturing using x-ray tomography images and the above mentioned image analysis program. For example, he calculated the average diameter of the machine direction and cross-machine direction polymeric fabric strands and then compared them to physical measurements using a microscope. He found out that the results from x-ray tomography and image analysis compared exceptionally well with less than 3 percent error. The image analysis system that they have developed uses a number of MSI resources, including the IBM SP supercomputer, with codes written in C++, MATLAB, etc. Jerome also organized the format making it suitable for scientific presentations.

Eric Johnson, majoring in Biomedical Engineering at the University of Minnesota worked with Professor William Gleason (Department of Laboratory Medicine and Pathology). Eric worked in two major areas: parallelization of CIDentify from the Lutefisk/CIDentify software package, and using a molecular dynamics simulation and visualization (NAMD/VMD) software package in interactive molecular dynamics simulations. The NAMD/VMD is used because of its interactive capability. It is possible for large-scale molecular dynamics calculations to be carried out by NAMD on the supercomputer, while visualization is accomplished on a local machine by VMD without the need for special graphics pipes. Not only can the physical results of the calculations be observed as they occur, the user can intervene by introducing forces or moving molecules and then immediately view the results. To investigate this software as a tool for computation and visualization Eric looked at Vascular Endothelial Growth Factor (VEGF). After becoming acquainted with the NAMD/VMD program, he learned to use the XPLOR software package to generate protein structure files from structures found in the protein database (PDB), obtained through x-ray crystallography and nuclear magnetic resonance spectroscopy. These two types of files, when properly created and modified can be combined with existing protein topology and parameter files to run dynamic simulations. Eric used this process to simulate the binding of VEGF to a receptor and to an antibody fragment. The dynamic simulation gives additional insight into the binding, supports the argument that the dynamic model created is sound, and the structure contained in the PDB accurately portrays the actual binding mode. Eric also worked on developing topology and parameter files for simulating the dynamic behavior of a heparin model.

Majoring in Computer Science at Columbia University, Benjamin Langmead worked with Professor Yousef Saad (Department of Computer Science and Engineering). Benjamin's work was geared toward enhancing and improving SPMATH, a package developed entirely with three undergraduate interns from MSI. SPMATH is a package to help solve sparse matrix problems (see also, our article Improving the Efficiency of Sparse Matrix Storage, page 6). Benjamin developed a new graphics interface based on the Total Command Language/Toolkit (TCL/TK). He also improved portability using Autoconf configure scripts, fixed a few computational bugs and improved functionality. As a result of Benjamin's work, this package has now been made available to the public (see: www.cs.umn.edu/~saad/software).

John OÕLeary from the University of Chicago.

John O'Leary, majoring in Computer Science and Biology at the University of Chicago, worked with Leonard Banaszak (Department of Biochemistry, Molecular Biology and Biophysics). John worked primarily on using genomic data to understand protein structures. Protein structure determines protein function, but how this occurs it not yet fully understood. John focused on using the large number of protein sequences in the National Center for Biotechnology's database as a tool to more clearly understand the role played by various parts of the structures of the enzymes in the tricarboxylic acid cycle. Primarily, John did a literature survey on both protein structure and common bioinformatics tools, and then ran sequence and structure databases to analyze the data. He used Perl programming to parse and manage the large amount of data that was generated through database searches. This resulted in the construction of a module that, after the entry of a protein name, makes multiple web requests in order to gather and separate out only the appropriate data for running a multiple sequence alignment. Multiple sequence alignments were determined using CustalX, and after their analysis with another Perl script, diagrams were made (using InsightII, RasMol, and Molscript) of the structures with conserved residues highlighted. Eventually, by using visualization software it can be discovered not only which residues are crucial to enzymatic function, but also why they are crucial.

Mariah Olson, majoring in Computer Science and Biology at University of Minnesota, Duluth worked with Professor David Levitt (Department of Physiology). Mariah worked on the development of a web-based database that could be used to carry out intelligent searches of the structures and sequences of antibodies and the peptides that they bind. She combined several different databases into one relational SQL database. This required extensive use of Perl scripts. Finally, Mariah began the development of an interactive web site to allow general use of these databases.

Sarah Betterman, Benjamin Langmead, Adam Butensky-Bartlett, and Nicholas Olson discuss one of the summer's tutorials.

Nicholas Olson, majoring in Electrical Engineering at University of Maryland, College Park, worked with Professor David Yuen (Department of Geology and Geophysics). Nicholas worked on a design map for data mining that could work over the Internet. This program would display an image developed from previous data, and, after placing a grid over the image, the program would provide the average, standard deviation and a histogram for each square within the grid. By enlarging or shrinking the grid the program will change the range of values in the histogram.

Jonathan Othmer, majoring in Mathematics and Music at Williams College (Williamstown, Massachusetts) worked with Professor Heinz Stefan (St. Anthony Falls Laboratory and Department of Civil Engineering). Jonathan joined Dr. Stefan's research group and did an analysis of a large stream temperature database for stream gauging stations throughout the United States. Initially, they looked for trends and patterns in the data relating to geographic distributions and record length. Then looking for a statistical upper bound on stream temperatures, they tried to improve upon previous work done on this topic. This is of interest to determine how far stream temperatures can rise if the climate warms. The physics of heat exchange suggest that stream water temperatures do not rise indefinitely. This is of great importance for stream water quality and biological habitat. Jonathan is continuing his work on a linearization method for estimated upper bounds on temperature. The best estimates of the upper bound for stream temperatures were presented in histogram and map format.

Rice University Computer Science major, William Ryan worked with Professor Thomas Jones (Department of Astronomy). William's work concentrated on two graphical user interfaces (GUIs) used by astronomy researchers. The first program is used to position the viewing space and setting the input parameters to a synthetic observation of a radio galaxy. William succeeded in having this program interface with the Fortran program, running the observation. The second GUI performs numerous operations on data generated by the previously mentioned synthetic observation program. Among these, the most complicated was the gathering of data along a line-of-sight through the two-dimensional image representation of three-dimension. This interprocess communication with existing Fortran code was then modified. William also did research analyzing the polarization of synthetic data, primarily to observe the orientation of magnetic fields, but also to look at Faraday rotation.

Zachary Garbow and Samuel Stechmann discuss their research while enjoying one of the program's luncheons.

Dawn Schafer, majoring in Biochemistry at Bethel College worked in the laboratory of Professor Ian Armitage (Department of Biochemistry, Molecular Biology and Biophysics). After an extensive literature review of nuclear magnetic resonance (NMR), Dawn worked on the basic theory and experimental aspects behind one- and two-dimensional NMR, using Varian software (6.1B VNMR) and NMR instrumentation to acquire, process and analyze data and create high-quality spectra. Dawn found that the most challenging aspect was associated with shimming or obtaining a homogeneous magnetic field around her sample. With the assistance of NMR facility personnel, Dr. Beverly Ostrowski, she learned to acquire one-dimensional, and the more challenging, two-dimensional spectra, specifically Correlation Spectroscopy, Total Correlation Spectroscopy, and Nuclear Overhauser Effect Spectroscopy on her protein sample.

Samuel Stechmann, majoring in Physics, Applied Mathematics and Chemistry at the University of St. Thomas worked with Professor Donald Truhlar and graduate student Ahren Jasper in the Department of Chemistry. Samuel worked on a time-uncertainty method for enabling classically forbidden electronic transitions in trajectory surface hopping calculations. He worked on a nonadiabatic trajectory (NAT) code that is written in Fortran and is used to simulate quantum mechanical photochemical reactions using semiclassical approximations. With such semiclassical approximations, the electrons are treated using quantum mechanics and the nuclei are dealt with using classical mechanics. The quantum mechanical electronic motion gives rise to potential energy surfaces over which the nuclei are propagated quasi-classically. Using this semiclassical approach simplifies the calculations, thereby reducing the expense of the simulation as compared to full quantum mechanics. Therefore, the development of these semiclassical methods is important for simulations on larger chemical systems for which quantum mechanical calculations are too expensive to carry out. They used the results (the electronically nonadiabatic transition probabilities and vibrational and rotational moment of the products) of quantum mechanical simulations on small (atom-diatom) systems as bases of comparison for the various prospective semiclassical methods.

Summer 2002 Undergraduate Internship Program

The Supercomputing Institute is pleased to announce its Undergraduate Internship Program for Summer 2002. Summer appointments will be full-time, ten-week appointments, and will run from June 10th through August 16th, 2002. A student interested in becoming an intern must be an undergraduate student during the internship to be eligible and must be a citizen or permanent resident of the United States or its possessions.

All applications are evaluated competitively based on the qualifications of the applicant and the availability of a suitable project. Applicants should review the research projects list and choose projects in which they are interested, although they may be offered other projects due to availability.

Further information, application forms, and project lists are available on the Internet at:


Application form and project lists are also available by contacting:

Undergraduate Internship Coordinator
University of Minnesota
Supercomputing Institute
1200 Washington Avenue South
Minneapolis, Minnesota 55415-1227

Phone: (612) 624-2330
Email: uip@msi.umn.edu