IDRE First Annual Kickoff Poster Proposals

Learning and Retention from an Online Tutorial among Resident Physicians

Authors:

Douglas S. Bell, MD, PhD; Jerilyn Higa, MS; Carol M. Mangione, MD, MSPH

Abstract

We randomized 91 residents who completed an online diabetes tutorial to take a post-test after 0, 1, 3, 8, 21, or 55 days. 87 subjects provided complete follow-up data (96%). Knowledge scores were 2.5 standard deviations above baseline for those tested immediately; gains were half as great after 8 days; no retention remained at 55 days. In linear regression modeling, critical appraisal skills and time spent on the interactive tutorial correlated with greater learning.


Return to Top

Computational Chemistry at UCLA

Author:

Ken Houk

Abstract

The Houk Group at UCLA applies the methods of computational chemistry to problems in organic and physical chemistry, pharmacology, and enzyme design. Quantum mechanical methods are used to calculate the energies of molecular structures involved in a reaction. Determination of a reaction mechanism requires the calculation of transition structures, energy maxima along a coordinate partway between reactants and product. This requires density functional theory (DFT), basis set extrapolation methods, coupled-cluster theory, quantum molecular dynamics, QM/MM, and time-dependent DFT. Numerous projects are underway to resolve mechanistic questions in the following areas:



Return to Top

Enzyme Design and Biological Catalysis

Author:

Ken Houk

Abstract

Designing enzyme active sites based on QM transition states of novel reactions and the surrounding protein residues needed for catalysis.



Return to Top

HNO and NO chemistry

Author:

Ken Houk

Abstract

These small molecules have central roles in the circulatory and respiratory system as well as signaling throughout the body.



Return to Top

Organometallic reaction mechanisms

Author:

Ken Houk

Abstract

Organometallic reaction mechanisms comprise important reactions in industry and natural product synthesis but are challenging to calculate accurately.



Return to Top

Pericyclic reaction mechanisms

Author:

Ken Houk

Abstract

Pericyclic reaction mechanisms, important to organic synthesis as bonds, sometimes many at once, are formed stereoselectively.



Return to Top

Stereoselectivity and organocatalysis

Author:

Ken Houk

Abstract

Stereoselectivity and organocatalysis with small organic compounds, such as proline, can lead to more efficient and greener synthetic routes.



Return to Top

Molecular Devices, Materials, and Host-Guest Chemistry

Author:

Ken Houk

Abstract

Supramolecular structures and extended conjugated systems can be used for molecular storage and delivery and nanoelectronics.



Return to Top

Hybrid Simulation of Ion Acoustic Waves Including Coulomb Collisions

Authors:

Richard Wang; Bruce I. Cohen; Russel E. Caflisch; Andris M. Dimits; Yanghong Huang; Giacomo DiMarco

Abstract

Kinetic simulation of collective phenomena including Coulomb collisions in inhomogeneous plasma presents significant multi-scale challenges. When the ratio of the collisional-mean-free-path of an ion or electron species to the local scale length of the plasma properties or the electromagnetic fields varies from very much greater than unity (kinetic limit) to very much smaller than unity (fluid limit) over a domain of interest, comprehensive simulation becomes difficult; and a brute-force, first-principles approach is typically impractical because of the severe computational stiffness of the underlying physics. This presentation reports progress on the development of a kinetic-fluid hybrid technique for plasma simulation intended to address such multiple scale situations. A specific application to the simulation of ion acoustic waves including both Landau damping and Fokker-Planck Coulomb collisions is presented. The hybrid approach described here is twofold. First of all, ions are assumed to be all particles initially and electrons are assumed to be fluid. Secondly, ions are further represented by the sum of a time-dependent Maxwellian distribution and a collection of discrete particles, which ultimately will reduce the computational burden. Numerical simulation of the hybrid method will be presented.


Return to Top

Efficient Association Study Design via Power-Optimized Tag SNP Selection

Authors:

Buhm Han; Hyun Min Kang; Myeong Seong Seo; Noah Zaitlen; Eleazar Eskin

Abstract

Discovering statistical correlation between causal genetic variation and clinical traits through association studies is an important method for identifying the genetic basis of human diseases. Since fully resequencing is prohibitively costly, genetic association studies take advantage of correlation structure between single nucleotide polymorphisms (SNPs) by selecting a subset of SNPs to be genotyped (tag SNPs). While many current association studies are performed using commercially available high-throughput genotyping products that define a set of tag SNPs, choosing tag SNPs remains as an important problem for both custom follow-up studies as well as designing the high-throughput genotyping products themselves. The most widely used tag SNP selection method uses a criteria based on the correlation between SNPs (r^2). However, tag SNPs chosen based on an r^2 criterion do not necessarily maximize the statistical power of an association study. We propose a study design framework that chooses SNPs to maximize power and efficiently measures the power through empirical simulation. Empirical results show that our method gains considerable power over the r^2-based method, or equivalently reduces the number of individuals required. Our power-optimized whole genome tag sets reduce the sample size to achieve 80% power by up to 35% compared to the currently available commercial chips. Our method is publicly available via web server at http://design.cs.ucla.edu.


Return to Top

Large video collections

Authors:

Francis Steen, Communication Studies; Tim Groeling, Communication Studies Otto Santa Ana, Chicano Studies

Abstract

Large video collections pose interesting problems for search and access. We have a collection of 30,000+ hours of television news, automatically indexed using timestamped closed captioning and extracted frames. Rather than relying on machine vision, we propose to design a system to leverage the human ability to rapidly extract patterns from a complex visual matrix.


Return to Top

A Hybrid Monte Carlo Method for Coulomb Collisions

Authors:

Russel E. Caflisch; Bruce I. Cohen; Giacomo DiMarco; Andris M. Dimits; Yanghong Huang; Richard Wang

Abstract

This poster describes a hybrid computational method for Coulomb collisions in a plasma that combines a Monte Carlo particle simulation and a fluid dynamic solver in a single uniform method throughout phase space. The new method is based on a hybrid representation of the velocity distribution function f(v), as a combination of a Maxwellian equilibrium M(v) and a collection of discrete particles g(v). The Maxwellian M evolves in space and time through fluid-like equations, and the particles in g convect and collide through Nanbu's Monte Carlo particle method. Interactions between M and g are represented by a thermalization process that removes particles from g and includes them in M and a dethermalization process that samples particles from M and inserts them into g. As test cases for the hybrid method, we have used relaxation of an anisotropic Maxwellian and evolution of a bump-on-tail.


Return to Top

The Saxon Cluster, the Open Science Grid and the Compact Muon Solenoid (CMS) Experiment at CERN

Authors:

Joan Slottow; Jordan Tucker; David Saltzberg

Abstract

This poster highlights the work of UCLA physics researchers who are participating in the Compact Muon Solenoid (CMS) Experiment at CERN's Large Hadron Collider. It discusses the science involved and the computational requirements of that science. The major focus is on the Saxon Cluster, the Open Science Grid of which it is a part, and the software that supports CMS research.


Return to Top

Ordering Properties of Intervals: an enabling perspective for cross-genome databases

Authors:

Alexander V Alekseyenko, UCLA School of Medicine; Christopher J Lee, UCLA, Department of Chemistry and Biochemistry

Abstract

Motivation: The exponential growth of sequence databases poses a major challenge to bioinformatics tools for querying alignment and annotation databases. There is a pressing need for methods for finding overlapping sequence intervals that are highly scalable to database size, query interval size, result size, and construction / updating of the interval database.
Results: We have developed a new interval database representation, the Nested Containment List (NCList), whose query time is O(n+log N), where N is the database size and n is the size of the result set. In all cases tested this query algorithm is 5 - 500 fold faster than other indexing methods tested in this study, such as MySQL multi-column indexing, MySQL binning, and R-Tree indexing. We provide performance comparisons both in simulated datasets and real-world genome alignment databases, across a wide range of database sizes and query interval widths. We also present an in-place NCList construction algorithm that yields database construction times that are approximately 100-fold faster than other methods available. The NCList data structure appears to provide a useful foundation for highly scalable interval database applications.
Availability: NCList data structure is part of Pygr, a bioinformatics graph database library, available at http://sourceforge.net/projects/pygr


Return to Top

Research Performed on ATS Hosted Clusters

Authors:

Prakashan Korambath; Joan Slottow

Abstract

Over the past 5 years ATS has hosted a number of clusters belonging to different researchers in its Data Center. This poster focuses on the research that is being done on these clusters.


Return to Top

The Hoffman2 Cluster -- a Shared Cluster System Under the IDRE Cluster Hosting Program

Authors:

Joan Slottow; Bill Labate

Abstract

The Cluster Hosting Program provides cluster hosting services to campus researchers in a way that effectively manages the limited high-end data center space on campus. It both maximizes the number of supported customers and minimizes the labor to support a given cluster, while at the same time providing a rich and robust set of hardware, software, application and support services. The Cluster Hosting Program balances the needs of research teams that need many closely coupled nodes as well as those that need support in housing and maintaining smaller clusters. Under the auspices of the Cluster Hosting Program, ATS is building the Hoffman2 Cluster, a new powerful compute cluster. This poster gives details about the Cluster Hosting Program. It highlights the Hoffman2 Cluster, its architecture and the researchers that are contributing virtual clusters to the Hoffman2 as a shared cluster.


Return to Top

Coarsening dynamics of Cahn-Hilliard equation and its application in road image inpainting

Authors:

Shao-Ching Huang; Andrea L. Bertozzi

Abstract

We use the Cahn-Hilliard equation as a model of coarsening systems to study its long time behavior under the influence of discretization errors. In particular, we investigate the scaling properties of discretization errors to guide the design of a gradient stable time stepping method, the ultimate goal being to compute the slow dynamics both efficiently and accurately. Some results of applying the Cahn-Hilliard equation to road image inpainting problems will also be presented.


Return to Top

Remote Visualization Resources

Author:

Scott Friedman

Abstract

We have undertaken an effort to extend the reach of powerful interactive cluster based visualization resources to researchers by making them available remotely. Because they are used interactively, visualization resources typically have a fixed location researchers must travel to in order to use. Often, this can be inconvenient even if the resource is available across a campus. By leveraging the latest generation high speed WAN technologies we aim to bring high performance interactive visualization to the researcher.

Accomplishing this required us to extend our existing twenty-four node Infiniband based visualization cluster with an 10G iWarp capable remote visualization bridge node. Remote users connect to this node establishing an interactive visualization session. The system is designed to support multiple simultaneous HD quality interactive visualizations by sharing the capabilities of the back-end cluster.

This work demonstrates a novel use of high performance network resources as well as a mixed interconnect topology.


Return to Top

The CNSI Electron Imaging Center for Nanomachines and High Performance Image Processing

Authors:

Z. Hong Zhou; Sergey Ryazantsev; Ivo Atanasov; Wong H. Hui; B. C. Regan; Jeff F. Miller; Leonard Rome

Abstract

Recent advances have made electron imaging an indispensable tool for determining the three-dimensional (3D) structures and molecular interactions of macromolecular complexes or biological nano-machineries. The newly established Electron Imaging Center for Nanomachines (EICN) at UCLA aims to provide this emerging technology in its finest forms to both nano biology and nano-materials science researchers. Two slightly different modalities of electron imaging – single particle cryo-electron microscopy (cryoEM) and cryo-electron tomography (cryoET) – are commonly employed to visualize or “see” nano-biological machineries or particles of different structural property. For nano-particles with a homogenous structural organization, such as protein/DNA/RNA complexes and viral capsids, cryoEM is used to record a low-dose image for particles embedded in vitreous ice. Images of thousands of randomly oriented “single” particles are then averaged to obtain a 3D structure to near-atomic resolution (0.3-0.6nm). For materials and complexes with pleomorphic or dynamic structures where averaging is not possible, cryoET is used to obtain their 3D structures at molecular resolution (2-5 nm) from a tilt image series of the samples. These structural methods have made possible for biologists, chemists and materials scientists to determine the 3D structures of a wide variety of nanometer-scale assemblies, devices and materials. However, due to the intrinsic high level of noise in cryoEM images, data processing and data management present some of the greatest computational challenges in modern biology. Current algorithms in cryoEM 3D reconstruction scale poorly, O(N3) to O(N4), when targeting high resolution. Most computation tasks in cryoEM are well fit to high performance distributed computing, as well as state-of-the-art graphics processing. Therefore, these computational challenges represent great opportunities for computational scientists for developing high-performance computing algorithms or strategies using parallel vector machines.


Return to Top

Parallel Particle-in-Cell Codes in Plasma Physics

Author:

Viktor K. Decyk

Abstract

Particle-in-Cell (PIC) codes, which integrate the trajectories of charged particles in the electromagnetic fields they generate, are widely used in plasma physics. The first parallel versions were developed on the Hypercubes at JPL in 1987. Advances in hardware and algorithms now allow problems as large as 12 billion interacting particles on a 1024 cubed mesh to be run on a 2000 CPU system. Parallelization strategies, issues in achieving high performance and the development of a general framework for PIC codes will be discussed.


Return to Top

The UCLA Grid Portal and the UC Grid Portal

Authors:

Joan Slottow; Prakashan Korambath; Kejian Jin

Abstract

The University of California has adopted the UCLA Grid Portal (UGP), a web interface to cluster services, to create grids of computational clusters at its 10 campuses. UGP provides a common way of working with computational clusters in a grid which is both secure and easy to use. Because it handles all grid related operations for users, UGP eliminates any requirement that users install and use client-side grid software or certificates and lets the users concentrate on their research. UCLA Campus resources can be accessed from the UCLA Grid Portal or accessed in conjunction with resources from other campuses at the UC Grid Portal. UGP is written mainly in Java as a set of portlets for the GridSphere Portal Framework and makes extensive use of Web 2.0 technologies throughout. Collaborators at other UC campuses have also contributed to the development, which is ongoing. In this poster we will show how easy the computational resources at UCLA are to use from the UCLA Grid Portal.


Return to Top

Sensitivities of stratocumulus organization to precipitation

Authors:

Verica Savic-Jovcic; B. Stevens

Abstract

In Arakawa's 1975 WMO report, stratocumulus are identified as one of the Canonical Cloud forms whose representation is essential to accurate simulations of the general circulation. In the intervening decades, work by himself, his students and collaborators has greatly advanced our understanding of this critical cloud regime. Using large- eddy simulations on large domains (25.5 x 25.5 x 1.5 km^3, with 512 x 512 x 97 grid points) we explore the effect of precipitation on stratocumulus structure and dynamics. In particular, we investigate the hypothesis that drizzle can induce a transition in cloud planform from closed cellular circulations, characterized by well mixed boundary layers with significant cloud cover, to open cellular circulations, characterized by decoupled boundary layers and less cloud cover. The simulations are shown to support this hypothesis and are consistent with available measurements. Additional agreement between the observations and our simulations is that stratocumulus can have long-lasting and locally intense drizzle, which is also associated with the increased horizontal variability in the temperature and moisture fields, but reduced horizontal variance of vertical velocity, as well as with pools of elevated equivalent potential temperature.


Return to Top

Advances in Particle Swarm Optimization (PSO) for Engineering Electromagnetics

Authors:

Nanbo Jin; Yahya Rahmat-Samii

Abstract

The particle swarm optimization (PSO) is a recently proposed evolutionary algorithm by applying the social-psychological metaphor in the natural swarming behavior. This poster presents the development and application of a versatile UCLA-PSO engine in a wide range of engineering electromagnetic optimization problems. The optimization kernel has different implementations (real-number, binary, hybrid, single- and multi-objective) according to the variable to be optimized, and it is proved to be effective in searching the global optimum in high-dimensional, non-linear solutions spaces. Representative examples include the design of correlator antenna arrays, aperiodic antenna arrays, metamaterial-based antennas and multi-layer planar material for RCS reduction, with both numerical simulation and measurement results presented to validate the high-quality performance obtained by applying the swarm intelligence.


Return to Top

A Novel Reflector Surface Distortion Compensation Technique Using a Sub-Reflectarray

Authors:

Shenheng Xu; Yahya Rahmat-Samii

Abstract

With the increasing interest in the application of large inflatable reflector antennas operating at high frequencies, the requirements on the reflector surface accuracy become more demanding. However, thermal and gravitational effects inevitably cause certain reflector surface distortions, and degrade the overall antenna performance. A novel reflector surface distortion compensation technique using a sub-reflectarray is thus proposed. A microstrip reflectarray is used as a subreflector, illuminated by a primary feed. The reflection phase of each sub-reflectarray element is properly designed, so that an additional phase shift is added to the re-radiated wavefront, which can be ultimately used to counteract the aperture phase errors caused by the main reflector surface distortions. One of the key steps in this design, the extraction of the required reflection phase of the sub-reflectarray, is extensively explored by two different methods. This technique presents several favorable features. The microstrip sub-reflectarray is low-profile, inexpensive, and easy-to-build. Only one primary feed is required, thus avoiding the complicated feeding network for array feeds. The added corrective apparatus is light-weighted, making this technique more attractive for spaceborne applications. And a reconfigurable design can be easily achieved if MEMS-controlled reflectarray elements are adopted.


Return to Top

A New Look at Phaseless Planar Near Field Measurements: Limitations Simulations, Measurements and a Hybrid Solution

Authors:

Farhad Razavi; Yahya Rahmat-Samii

Abstract

There is a growing interest in utilizing the phaseless measurements for characterizing the antennas patterns. These measurements are cost effective, well-adapted to higher frequencies, relax the probe positioning error and do not need phase stability. In phaseless measurement techniques, one attempts to construct the phase distribution based on some known amplitude information or constraints. This process is usually referred to as “phase retrieval”. There are a number of numerical methods for solving this problem. A famous method in this arena is called Iterative Fourier Technique (IFT). This is a nonlinear algorithm which has many promising capability like the usage of FFT (for acceleration of iterations) and the appropriateness for planar measurements. Unfortunately there are some challenges involved in applying IFT for general complex patterns, like scanned beam antennas. This study has proposed a solution to this problem by introducing a searching mechanism based on the Differential Evolutionary Algorithm which optimizes the initial guess for IFT algorithm. The effectiveness and applicability of this method is shown through both simulation and measurements.


Return to Top

Fast ignition fusion research using particle- in-cell methods

Author:

John W. Tonge


Return to Top

An exact many-electron ab initio simulation method for bond-breaking reactions in liquids

Authors:

William J. Glover; Benjamin J. Schwartz, UCLA Department of Chemistry & Biochemistry

Abstract

Chemical reactions involve the formation and breaking of bonds between atoms and much of the field of chemistry is concerned with how chemical reactions progress. We are interested in a particular class of reactions that are initiated by light: condensed-phase photodissociation reactions. To study these systems we have developed a new simulation method that is able to solve Schrödinger's Equation (SE) exactly for the pair of electrons involved in the breaking chemical bond in an explicitly modeled solvent. By expressing the electrons' wavefunction on a 16^6 six-dimensional real-space grid we allow flexibility in the electrons' wavefunction to respond to the disorder in the liquid state. In this representation, solving SE involves diagonalizing an ~8 million by 8 million Hamiltonian matrix. Using the Jacobi-Davidson algorithm and fast Fourier transforms with an efficient representation of the electron-electron Coulomb operator we have achieved timings of tens of minutes for finding the lowest eigensolutions of the large Hamiltonian matrix. This allows us to use our method in molecular dynamics simulations where we diagonalize the SE at every time step. We present preliminary simulation results showing the influence of a solvent on the bond of a sodium dimer molecule.


Return to Top

 

webmaster@idre.ucla.edu