Confirmed Speakers

Keynote Speaker: Dr. Russ Miller

Distinguished Professor, University at Buffalo

http://www.cse.buffalo.edu/faculty/miller/

In this talk, we present an overview of the Center for Computational Research (CCR), the Cyberinfrastructure Laboratory (MCIL), and our research efforts in molecular structure determination, the results of which were listed on the IEEE poster “Top Algorithms of the 20th Century”. We present a framework that can be used to guide cyber-related activities in order to 1) accelerate discovery and comprehension, 2) create links between enabling technologists and disciplinary users, 3) create new techniques, algorithms, and interactions that improve efficiency of knowledge-driven applications in myriad disciplines, 4) enhance virtual organizations, 5) provide increased education, outreach, and training, and 6) enhance and expand relationships between academia and the corporate world.

We give an overview of our recent activities that fit within the aforementioned framework, including developments in enabling technologies, as well as supporting efforts in science, engineering, media, life sciences, transportation, urban planning, and humanities. We present an overview of our design and implementation of the New York State Grid (NYS Grid), including our efforts in grid monitoring, predictive scheduling, grid-enabling application templates, backfill detection and optimization, data repositories and operations, as well as in dynamic and automated allocation of resources.

Finally, we present an overview of our Shake-and-Bake method of molecular structure determination and its implementation on the NYS Grid.

Bio: Dr. Miller is UB Distinguished Professor of Computer Science and Engineering at SUNY-Buffalo, Head of the Cyberinfrastructure Laboratory, Founding Director (1998-2006) of SUNY-Buffalo’s High-Performance Computing Center, and Senior Research Scientist at the Hauptman-Woodward Medical Research Institute.

Dr. Miller founded the Center for Computational Research (CCR) at SUNY-Buffalo, where he served as Founding Director from 1998-2006. This worldwide leading supercomputing center provided key infrastructure to enable 21st century discovery and innovation. In addition to computational science and engineering, CCR supported many non-traditional areas, typically supporting hundreds of faculty-led projects per year.

The work presented in this talk was supported by NSF, NIH, DOE, NYS, the Oishei Foundation, the Wendt Foundation, NIMA, Dell, IBM, and HP. This talk represents joint work with numerous students, staff, and colleagues, who are listed in the talk.

Edmund B. Webb III

Associate Professor, Mechanical Engineering & Mechanics, Lehigh University

http://www.lehigh.edu/~inmem/webb.html

Atomic Scale Models and HPC: An Insider's Perspective
Since the first calculations on neutron diffusion in the late 1940s, numerical simulation has emerged as a third pillar of science and engineering research. Cast alongside theory and experimentation, numerical simulation now represents a significant body of research, across myriad disciplines. More specifically, those early diffusion calculations sparked the dawn of atomic scale numerical simulation. While atomic scale models were long relegated to the realms of academic discovery, advances in computing technology – including high performance computing – have pushed them into the realm of industry production. Nonetheless, the short timeframe associated with typical design to product cycles often precludes the use of such fundamental scale calculations. Benefits that might be garnered from more nuanced understanding are cast aside in response to a basic need of industry to get an operating product to market.

In this talk, a brief history and description of atomic scale modeling will be presented. Examples will be discussed of fundamental scientific studies one can perform with a freely available, open source, massively parallel computer code. While examples presented will focus on one realm of science – materials science and thermomechanical behavior – an attempt will be made to highlight some roles that atomic scale numerical simulation plays in current industry practices. The talk will conclude with a discussion of existing limitations and future possibilities.

Bio: Edmund Webb III received his doctorate in Ceramic and Materials Science and Engineering from Rutgers, The State University of New Jersey where he used atomic scale models to study the surfaces of oxide glasses. He then assumed a post-doctoral research position at Exxon Research and Engineering, applying atomic scale simulations to study hydrocarbon upgrade processes for automotive lubricant production. Following this, Prof. Webb joined the ranks of Sandia National Laboratories, where he worked for 12 years before coming to Lehigh University in 2010. As a national laboratory research scientist, Prof. Webb applied high performance computing resources to a range of materials and mechanics problems, including capillary driven fluid flow, friction mitigation, stress evolution in thin films, nanoscale thermal transport, and liquid droplet impacts.

Xiaolei Huang

Assistant Professor, Computer Science & Engineering, Lehigh University

http://www.cse.lehigh.edu/~huang/

A Parallel Cellular Automata on the GPU for Interactive Brain Tumor Segmentation
We present a novel method for 3D brain tumor volume segmentation based on a Parallel Cellular Automata framework, implemented on the GPU. Given the requirement of high accuracy and reliability in brain tumor segmentation, interactive methods are favored over fully automated ones. For an effective interactive segmentation, speed and usability are crucial. The Cellular Automata (CA) method supports 3D, multiple-label segmentation where the number of labels does not increase computational time or complexity. It is an iterative method where each cell independently follows a set of rules, naturally lending itself to an efficient parallel implementation. Further, the iterative nature of this method enables the user to interact with the method and visualize results at any time during the segmentation process and to correct mistakes in local areas on the fly. Our proposed CA method also introduces prior knowledge about labels gathered from user interactions as well as the brain atlas prior to incorporate global constraints into CA's local updating rules and improve accuracy. Exploiting the inherent parallelism of our algorithm, we adopted this method to the GPU and was able to compute segmentations nearly 45x faster than conventional CPU methods, enabling user feedback at interactive rates.

Bio: Xiaolei Huang received her B.E. Degree in Computer Science from Tsinghua University, China in 1999, and her M.S. And Ph.D. degrees in Computer Science from Rutgers University in 2001 and 2006, respectively. She is currently an Assistant Professor in the Computer Science and Engineering department at Lehigh University, Bethlehem, PA. Her research interests are in the areas of biomedical image analysis and computer graphics, focusing on segmentation, registration, and deformable model based methods.

Jeremy Hylton

Senior Staff Software Engineer, Google, New York

http://www.python.org/~jeremy/

Bio: Jeremy Hylton joined Google as a software engineer in 2004. Currently, he is based in Google's New York office, where he leads the search quality team, including the realtime search team. Prior to joining Google, Jeremy was a Python and Zope developer at Zope Corp, BeOpen PythonLabs and the Corporation for National Research Initiatives (CNRI). He received his bachelor's and master's degrees in computer science from M.I.T.

Gabriel Tanase

IBM T.J. Watson Research Center

http://parasol.tamu.edu/people/gabrielt/

The Anatomy of a Parallel Programming Language
Highly parallel computer architectures are increasingly available, and the peak performance of TOP500 computers still follows Moore's law. However parallel programming is difficult, and currently there is a large number of projects trying to address programmers productivity while developing large scale parallel applications.

In this talk we look at the various components of a modern parallel programming language: high level programming abstractions, necessary compiler support, runtime system and communication library. For each of these we discuss existing work and possible research directions. We use our own project - Unified Parallel C (UPC) for our analysis but the discussion is general and apply to any other parallel programming language or library.

Bio: Gabriel Tanase is a Research Staff Member at IBM T.J. Watson Research Center where he works on run time systems for parallel programming languages. He graduated with a PhD in Computer Science from Texas A&M University. His PhD work is on parallel data structures in the context of a novel C++ parallel programming library called STAPL. He received his Bachelor of Science from the Polytechnic University of Bucharest, Romania in 1999 and Master of Science from the same University in 2000. His research interests are in the area of high performance computing including parallel programming languages and libraries, parallel algorithms and generic programming.

Yana Bromberg

Rutgers University, Dept. Biochemistry and Molecular Biology

http://aesop.rutgers.edu/~dbm/ybromberg.html

Identifying protein functional sites using in silico mutagenesis.
The ability to identify functionally non-neutral non-synonymous SNPs (single nucleotide polymorphisms) may be used to guide experimental study of detrimental mutations, as well as substitutions that increase the fitness of particular phenotypes. This reasoning has spawned a number of computational approaches for identifying functional effects of nsSNPs. One such method, SNAP (Screening for Non-Acceptable Polymorphisms), utilizes only sequence-derived features and produces a well-calibrated reliability score for each prediction. SNAP’s ubiquity in application to all proteins, its accuracy, and its scoring system have opened another door in the study of function: in silico mutagenesis. Experimental mutagenesis studies are used for a wide range of biological reasons including identification of functionally important protein residues. Here we demonstrate that SNAP’s predictions can be used to mimic and augment experimental designs of mutagenesis studies and, in a cheaper and more efficient manner, identify protein functional sites.

Nurit Haspel

Assistant Professor in the Computer Science Dept.
at University of Massachusetts, Boston

Affordable Departmental Supercomputer facilitates the conformational modeling and simulation of protein dynamics.
In this work we apply large scale computation to explore the conformational space of short peptide sequences known to generate functional amyloids. This work is part of our efforts to develop a computational, theoretical and experimental framework to rationally design nano- and micro-structures made of amphiphilic hybrid materials which combine peptides used in the formation of amyloids with polyesters. Such exploration requires the use of massive computational resources in order to obtain a good description of the peptide conformational space. This exploration was made possible and efficient by the use of the NAMD molecular dynamics software installed on the Symmetric Computing Supercompuer, which allowed us to perform the research much more efficiently than any other locally available resource and without resorting to large national resources. My research lies in the areas of computational structural biology and structural bioinformatics. My goal is to better understand the structure and flexibility of proteins, to model conformational changes in proteins, and to design novel nano-structures which may contribute to the understanding of the self-assembly properties of proteins and facilitate experimental nano-design. In my research, I focus on both the development of novel algorithms and the application of state-of-the-art existing methodologies to various problems in molecular biology, nanobiology and biochemistry. A detailed description of my past and present research can be found here.

Bio: Nurit Haspel, Assistant Professor at UMass Boston: Nurit Haspel received her BSc, MSc and PhD from Tel Aviv university in Israel. She later did a 2 year postdoctoral work at the Department of Computer Science at Rice University in Houston, TX and in 2009 she joined the department of Computer Science at UMass Boston as an assistant professor. Her research area is structural bioinformatics - the application of computational methods to solving key biological problems. Specifically, she develops and applies computational algorithms based on concepts taken from computational geometry, graph theory and robotics to model the structure, function and dynamics of proteins and biomolecular interactions.

http://www.cs.umb.edu/~nurith/

Willy Wriggers

D.E. Shaw Research / Weill Cornell Medical College

http://biomachina.org/wriggers.html

Emergent complexity of multiscale computational modeling
Proteins and protein assemblies exhibit "emergent behavior" in both the temporal and spatial domains. As observed in long-timescale simulations or in models of large-scale systems, the behavior that emerges on these scales is not only more than the sum of the (temporal or spatial) parts, but quite different and unexpected. I will discuss two examples of modeling and refinement of protein structures in which coupling across time or spatial scales yields complex phenomena one cannot predict from isolated degrees of freedom: (1) slow conformational changes enabled by fast motion in a long molecular dynamics simulation of bovine pancreatic trypsin inhibitor (BPTI); and (2) simultaneous fitting of multiple atomic fragments into low-resolution data from electron microscopy. I will argue that, in the future, it may be useful to employ hybrid modeling and analysis techniques that permit the study of emergent phenomena in both the temporal and spatial domains.

Willy Wriggers develops statistical methods for understanding biomolecular systems through simulations. He is a full-time member of D.E. Shaw Research and also holds an academic appointment as Associate Professor at Weill Cornell Medical College. He came to New York from the University of Texas Health Science Center, Houston, where he was Associate Professor of health informatics and molecular medicine. His research interests are time series analysis, multi-scale modeling, molecular imaging, and computational geometry. His scientific career has also included a faculty position in the Department of Molecular Biology at the Scripps Research Institute (TSRI), and postdoctoral positions in both electron microscopy (Department of Cell Biology, TSRI) as well as theoretical chemistry (University of California, San Diego). Willy earned a Ph.D. in physics from the University of Illinois, Urbana-Champaign for his work on molecular dynamics simulations of biomolecular machines.