Title: Assessing the Global Impact of Histone Acetylations on Gene Expression in Yeast

Abstract: Gene activities in eukaryotic cells are regulated by the concerted action of transcription factors and the chromatin structure. Saccharomyces cerevisiae (budding yeast) contains at least 14 lysines residues at four different core histone proteins that can be acetylated. The combined effect of various acetylation sites is still poorly understood. In this talk, I will present an integrative approach combining information from acetylation, nucleosome occupancy, gene expression, and sequence information to systematically investigate the regulatory code of histone acetylation in Saccharomyces cerevisiae.

The talk is based on joint work with Guo-Cheng Yuan, Wenxuan Zhong and Jun S. Liu.

Title: A revised view of the library of expressed mammalian genes

Brendan J. Frey
University of Toronto and Canadian Institute for Advanced Research

Abstract: In the past 10 years, researchers have been trying to get a clear picture of how many protein-coding mammalian genes have yet to be discovered. Recent DNA microarray and SAGE experiments suggest that an unexpectedly large proportion of the genome is transcribed and may code for proteins. Using DNA microarrays and a carefully designed engineering tool based on artificial intelligence techniques, we have produced a revised view of the library of mammalian genes and their exons. In our work (published in Nature Genetics in Sep 2005), we survey the expression of 1.14 million DNA sequences from across the entire mammalian genome, in 37 different natural tissues. We find thousands of new transcribed DNA regions. However, our final conclusion is surprising and both closes a major chapter of genomics research and opens up several new avenues to explore.

Work done in collaboration with TR Hughes, QD Morris, N Mohammad, W Zhang, MD Robinson, S Mnaimneh, O Shai, R Chang, Q Pan, E Sat, J Rossant, BG Bruneau, JE Aubin and BJ Blencowe.

Biography: Brendan J. Frey is a faculty member in Electrical and Computer Engineering and the Banting and Best Department of Medical Research, at the University of Toronto. He was born on August 29, 1968, in Calgary, Alberta near the foothills of the Rocky Mountains, where he enjoyed hiking and camping with his family. In 1979, he started writing computer programs, attaching sensors to his home computer, and building simple robots. His university education was in the areas of physics, engineering and computer science, culminating with a doctorate from Geoffrey Hinton's Neural Networks Research Group at the University of Toronto. From 1997 to 1999, Frey was a Beckman Fellow at the University of Illinois at Urbana-Champaign, where he continues to be an adjunct faculty member in Electrical and Computer Engineering. From 1998 to 2001, he was a faculty member in Computer Science at the University of Waterloo. Currently, Frey is head of the PSI-Group at the University of Toronto. He has received several awards, given over 80 invited talks and published over 100 papers on advanced computer algorithms, molecular biology, computer vision and iterative decoding. More information is available from his research group's website, http://www.psi.toronto.edu.

Title: Bio-Pathway Logic

Abstract: The talk will discuss a style of high-level bioinformatics models of cell biology that can be used to study biological pathways. The models are based on a simple logic called rewriting logic which is efficiently implemented and supported by formal analysis tools in the Maude language. These models are very natural, because a biochemical reaction that may have for example a textbook description of the form:
     A B -C D
can be simultaneously understood as a rewrite rule, so that cell dynamics is modeled as concurrent rewriting. I will explain how these models are then executable and analyzable. The talk reports on joint work with colleagues at SRI International and focuses on the essential ideas. Towards the end I will describe further developments and future possible generalizations of the model.

Biography: Jose Meseguer is Professor of Computer Science at UIUC. His main interests are in Formal Methods and Semantics of Computation. He proposed rewriting logic in 1990 and has led the design of the Maude language. He has published about 200 scientific papers, many of which are highly cited in the literature.

Title: e-Photosynthesis - A system biology platform for guiding improvement in photosynthesis and crop productivity (research with Eric de Sturler and Steve P. Long)

Abstract: Theoretical analysis has shown that photosynthesis is the only major route for further improvement of crop productivity. Potential targets for improving photosynthesis include modification of photosynthetic metabolisms, anatomical structures of photosynthetic apparatus, and canopy structures. Given thousands even millions of permutations to test, identifying targets to improve photosynthesis using experimental approach is time consuming, costly, and impractical. The e-Photosynthesis research platform aims to integrate the genomic, proteomic, and metabolomic information, together with the microscopic structure of leaf and the macroscopic structure of canopy to predict the efficiency of photosynthesis under different environmental conditions for different natural or genetically modified plants. It can also be used to test hypothesis related to photosynthesis. The presentation will describe the basic structure of e-Photosynthesis, the methodology used to build e-Photosynthesis, supporting algorithms for apply e-Photosynthesis in research.

Bio: Xinguang Zhu received his MS degree from Institute of Botany, Chinese Academy of Sciences in 1999. He received his Ph.D. from UIUC in 2004 with a thesis titled "Computational Approaches to Guiding Biotechnological Improvements in Crop Photosynthetic Efficiency." His main research interest is to accurately simulate the photosynthetic and associated metabolisms in natural environmental conditions at different time scales, study the system properties of the photosynthesis system, and explore ways to increase photosynthesis yield. He is especially interested in combining the large amount of genomic, microarray, proteomics, and metabolomics data with modeling techniques to improve understanding about photosynthesis.

Title: Exploiting the Domain Structure for Gene Recognition from Biomedical Text

Abstract: Automatic recognition of gene names and protein names is a critical first step towards information extraction and text mining in biomedical literature. The state-of-the-art biomedical named entity taggers are mostly based on supervised learning methods, and their performance depends largely on the availability and the quality of the training data. Because of the different gene naming conventions among different biological species, the performance of a gene tagger trained on text containing gene names of certain species may degrade significantly when it is used to recognize gene names of new species. We propose an approach to gene recognition (and to named entity recognition in general) that exploits the domain structure in the training data to improve the performance of the tagger on new domains. We show that by combining the feature ranking in each training domain and imposing such generalizability-based new feature ranking as a prior, we can learn patterns that generalize well in all domains. Our results from three sets of experiments show that the proposed method outperforms a baseline method that does not consider the domain structure in the training data.

Biography: Jing Jiang is a third-year Ph.D. student in the Department of Computer Science at UIUC. Her research interests are mainly in information extraction and its application in biomedical literature.

No seminar meeting this week, due to the CS/IGB Bioinformatics Summit on March 6, 8, and 9.

Streaming videos and slidesets from the Bioinformatics Summit are here.

Title: Automated Natural Language Processing for Data Integration and Curation in the Wnt Signaling Pathway

Streaming video: [Windows Media format]

Abstract: We have developed a natural language processing system that is able to identify references to biological interaction networks in free text and automatically assembles a protein association and interaction map. The pipeline is fully automated, and derives key Wnt-pathway associated proteins and biological names from the literature itself using a chi-squared analysis of noun-phrases overrepresented in this literature with respect to the general signal transduction literature. Using identified terms which were over- represented in the Wnt literature with respect to MeSH-annotated signal transduction papers, we formed a base named entity dictionary to which we then appended full-parse derived protein names, and a gold standard set of proteins annotated by a Wnt-signaling review collection. The dictionary served as a name list for a full exhaustive assertion extraction step on the corpus which yielded annotations involving key Wnt-related molecules which were missing or different from those in the canonical diagram, but are described by the literature. Our results suggest that software exploiting a combination of NLP techniques for information extraction could form a valuable first-pass tool for assisting human annotation and maintenance of signal-pathway models.

Bio: Carlos Santos is a Ph.D. student at the University of Michigan's Bioinformatics Program. He works in the Natural Language Processing group under the supervision of Dr. David States in collaboration with the National Center for Integrative Biomedical Informatics. His current research interests are in applying natural language processing techniques to better understand and integrate facts and data from the biomedical literature with existing bioinformatics databases in order to better understand complex disease processes.

He completed his master's degree in Bioinformatics at the University of Michigan in the winter of 2003, as well as completed an undegraduate degree in Computer Science from Washington University in St. Louis in 2000. At Washington University, he collaborated with Dr. David States' lab at the Institute for Biomedical Computing.

Title: Modeling Dynamical Biological Systems:  DNA Helices, Semistructured Proteins, and Stem Cell Populations

Background:  Mike Colvin is Professor in the School of Natural Sciences at the University of California at Merced, a new campus that admitted its first graduate students in the Fall of 2004 and undergraduates in 2005.  Prior to going to Merced, Professor Colvin was head of the Computational Biology Group in the Biology and Biological Research Program at Lawrence Livermore Laboratories.  At Merced, Colvin is Director of the University of California-Merced Center for Computational Biology (UCM-CCB).

The UCM-CCB is a research and education center at the first new research university of the 21st century, the University of California in Merced, California. The UCM-CCB sponsors multidisciplinary scientific projects in which biological understanding is guided by computational modeling. The Center also facilitates the development and dissemination of undergraduate and graduate course materials based on the latest research in computational biology. Under Mike Colvin's leadership, the Center is undertaking a number of activities that aim to recast biology as an information science:

• Hosting multidisciplinary research projects in computational and mathematical biology that will provide a rich environment for graduate and undergraduate research.
• Developing new mathematical and computational methods that are widely applicable to predictive modeling in the life sciences.
• Developing and disseminating computational biology course materials that translate new research results into educational resources.
• Extending the successes in achieving these objectives to other universities and "university-feeder" institutions such as community  colleges.
  

Title: Inference for Quantile Regression Models with Applications to GeneChip Data

Presenter: Huixia Wang
Department of Statistics
University of Illinois at Urbana-Champaign

Abstract: The traditional inference for the linear mixed models depends strongly on the normality assumption, which is easily violated in some applications. We develop a robust rank score test for linear quantile models with a random effect. The rank score test can be carried out at a single quantile level or jointly at several quantile levels. It is derived for homoscedastic error models, but is valid for inference on treatment effects in an important class of mixed models with heteroscedastic errors.

The proposed test is motivated by studies of GeneChip data to identify differentially expressed genes through the analysis of probe level measurements. We propose a genome-wide adjustment to the test statistic to account for within-array correlation, and demonstrate that the proposed test is highly effective even when the number of arrays is small. Our empirical studies of GeneChip data show that inference on the quartiles of the gene expression distribution is a valuable complement to the usual mixed model analysis based on Gaussian likelihood.

Title: Enhancing Access to the Bibliome:  The TREC Genomics Track

Streaming video: [Windows Media format]

Slide set: [Adobe PDF format]

Abstract: New “high throughput” technologies have led to a fundamental change in the conduct of biomedical research.  These technologies, exemplified by the gene expression microarray, generate substantial amounts of data.  The discovery that several genes are differentially regulated in a disease process generates a need for the researcher to seek information about those genes, their products, and the biological processes in which they are involved.  These information needs require us to build better information retrieval (IR) systems to search the biomedical literature and other databases.  This talk will describe an initiative whose goal is to build better IR systems through the development of standardized test collections that allow comparative evaluation of systems and algorithms with a common data set.  The talk will provide an overview of the Text Retrieval Conference (TREC) and describe the methods and results of the Genomics Track within it.

Bio: William Hersh, M.D., is Professor and Chair of the Department of Medical Informatics & Clinical Epidemiology at Oregon Health & Science University (OHSU) in Portland, Oregon, USA.  Dr. Hersh is a leader in the IR field, both within medicine and generally.  He has published over 100 scientific papers as well as the book, Information Retrieval:  A Health and Biomedical Perspective (Springer-Verlag, 2003).  Dr. Hersh also serves as Co-Editor-in-Chief of the journal Information Retrieval and is a member of the TREC Program Committee.  In addition, he directs the graduate educational programs in biomedical informatics at OHSU.  More information can be found about him on his Web site at www.billhersh.info.

Dr. Hersh is a native of Chicago, Illinois.  He received his Bachelor of Science in General Biology from the University of Illinois at Champaign-Urbana in 1980, followed by a Doctor of Medicine (M.D.) degree from the University of Illinois at Chicago in 1984.  After completing a three-year residency in internal medicine at University of Illinois Hospital in Chicago from 1984-1987, he did a three-year postdoctoral fellowship in medical informatics at Harvard University from 1987-1990.  He then joined the faculty of the School of Medicine at OHSU, where he has been the catalyst behind the development of the institution’s academic medical informatics program.

Title: Protein Structure Comparison from Line Weavings

Abstract: Proteins provide a rich domain in which to test theories of shape similarity. Sometimes the detection of common local structure is sufficient to infer global alignment of two proteins; at other times it provides false information. Proteins with very low sequence identity may share large substructures, or perhaps just a central core. There are even examples of proteins with nearly identical primary sequence in which alpha-helices have become beta-sheets.

The thesis of this talk is: Protein similarity detection leads naturally to algorithms operating at the metric, relational, and isotopic scales. Our work introduces a definition of similarity based on atomic motions that preserve local backbone topology without incurring significant distance errors. Similarity detection then seeks rigid body motions able to overlay pairs of substructures, each related by a substructure-preserving motion, without necessarily requiring global structure preservation. This definition is general enough to span a wide range of questions: One can ask for full rearrangement of one protein into another while preserving global topology, as in drug design; or one can ask for rearrangements of sets of smaller substructures, each of which preserves local but not global topology, as in protein evolution.

Bio: Michael Erdmann is interested in robots and proteins. He is a Professor of Computer Science and Robotics in the School of Computer Science at Carnegie Mellon University, and Associate Faculty in the Department of Computational Biology in the School of Medicine at the University of Pittsburgh. On the robotics side he is interested in the mechanics of manipulation and in task-level planning under uncertainty. On the biochemical side he is interested in the geometry of proteins and in algorithms for determining protein structure from sparse NMR data. Dr. Erdmann is an IEEE Fellow, a member of the Editorial Board of the International Journal of Robotics Research, and a former Presidential Young Investigator. In his spare time, Dr. Erdmann likes to travel, particularly to his home in the Pacific Northwest.