Dr. Kirk E. Jordan (Bio)
Emerging Solutions Executive
IBM Strategic Growth Initiatives/Deep Computing
Title: Is Blue Gene a System for Computational Biology?
Authors: Kirk E. Jordan, IBM Deep Computing
Abstract: Computation is playing an ever increasing and vital role in biology creating demand for new machines. Vendors strive to meet demands with advanced computer architectures such as IBM’s Blue Gene machine. In this talk, I will give a brief overview of the Blue Gene, architecture. I will present some results on a few example problems run on Blue Gene related to computational biology. These examples will give some view of scaling. I will then describe how one might approach using the Blue Gene System in a less traditional way. A challenge to the computational biology community will be to think of “Big” science problems with impact on society that until now or in current implementations have fallen short of the mark. Finally, I will mention opportunities that exist for the community to get access to Blue Gene.
David Osguthorpe, Ph.D.
Associate Professor and Director of the Computational Biology Division
Department of Pharmacology
University of Colorado Health Sciences Center
Title: The Differential Dielectric Model: A New View of Protein Folding
Abstract: Over the past few years I have been developing an ab initio
protein folding algorithm which does not make use of
statistical potentials but attempts to re-create the physics
of protein folding in a simplified geometry model. From this
work has developed a model of protein folding which I have
called the differential dielectric model. This model provides
an independent basis for the creation of the non-polar core of
proteins entirely separate from the classic hydrophobic
effect. It is also a component of the physical basis for the
classic bioinformatics-derived secondary structure
preferences. The physical rational for this model will be
described and its implications for protein structure and
Arizona State University
Challenges in Improved Sensitivity of Quantification of PET Data for
Alzheimer's Disease Studies
Rosemary A. Renaut and Hongbin Guo
Our recent research is concerned with the development of techniques for the> estimation of parametric data from dynamic positron emission tomographic> (PET) brain images. Improved sensitivity in quantification of the metabolic processes in the brain will be useful for medical studies of disease progression and drug impacts in patients with Alzheimer's Disease. While PET is a useful diagnostic tool due to its ability to image biological functions in vivo, unlike other imaging modalities, dynamic PET data suffers from low count statistics, leading to poor signal-to-noise ratios. This complicates the solution of the inverse problem for estimation of kinetic parameters which in our work describe the dynamics of the tracer, fluoro-deoxy-glucose (FDG). Our analysis brings together several numerical techniques, from thecompartmental model describing the biological function, to the inverse problem which this generates for the parameter estimation, to techniques for image restoration of very poor data, combined with both a robust statistical validation of the entire process and practical and efficient design of the underlying method. In this presentation I review some of the challenges associated with generation of global parametric image, and then focus on recent work for simultaneous estimation of the input to the compartmental model.