Peter Bajcsy will speak on "DNA Microarray Image Processing" on Feb. 4. Dr. Bajcsy is a research scientist with the Automated Learning Group in NCSA and an adjunct assistant professor in CS and ECE.

Presentation slides: [PDF]

The content of the talk will be based on one journal paper and two book chapters. The journal paper: "Bajcsy P. “GridLine: Automatic Grid Alignment in DNA Microarray Scans,” IEEE Transactions on Image Processing, VOL 13, NO 1, pp.15-25, January 2004." can be downloaded from the following URL: http://alg.ncsa.uiuc.edu/do/documents/publications

The book chapters:
(1) Bajcsy P., L. Liu and M. Band, “DNA Microarray Image Processing,” Chapter of the book "DNA Array Image Analysis: Nuts&Bolts" by Gerda Kamberova, Ph.D. (Ed.) published by DNA Press (in press).
(2) Bajcsy P., J. Han, L. Liu and J. Young, “Survey of Bio-Data Analysis from Data Mining Perspective,” Chapter 2 of Jason T. L. Wang, Mohammed J. Zaki, Hannu T. T. Toivonen, and Dennis Shasha (eds.), Data Mining in Bioinformatics, Springer Verlag, 2004, pp.9-39.
He will make copies available upon request.

Topic: "Protein structure prediction and design via rotamer libraries"

Mr. Wei Xie is a doctoral candidate under the supervision of Professor Nikolaos V. Sahinidis in the Department of Chemical and Biomolecular Engineering at the University of Illinois at Urbana-Champaign.  His research interests are in computational biology.

Abstract:  Protein structure prediction and design are fundamental problems in the understanding of life and drug design.  Solution of these problems requires sophisticated computational techniques.  Methods based on rotamer libraries have been recently applied to these problems with very encouraging results.  This talk will begin with a tutorial on prediction of protein structures through homologs, conformational analysis of protein side chains, and rotamer libraries.  We will next review several mathematical formulations for library-based approaches, computational complexity results, and algorithms. The emphasis will be on the various Dead-End-Elimination and Bounding Elimination criteria, which have, by far, been the most successful methods.  In the last part of this talk, we will present some of our results, including the tightness of a Linear-Programming-based lower bound, a Benders decomposition to solve very large-scale Linear Programming relaxations, and a new elimination criterion based on reduced costs that are computed via a combinatorial algorithm.

Dr. Gustavo Caetano-Anolles, associate professor of bioinformatics, UIUC Department of Crop Science, will speak on the evolution of the protein world.

Title: "Charting the living world: Structural phylogenomics"

Presentation slides: [PDF]

Abstract: The recent genomic revolution has resulted in massive acquisition of nucleic acid sequences, with over 180 completely sequenced genomes yielding about a million protein sequences. This effort outpaces structural genomics with its 24,000 structural entries. However, a small repertoire of protein architectures (known as protein folds) can be mapped onto about half of amino acid residues encoded in genome sequences. Consequently, the world of protein molecules, though uncharted, appears finite and its study feasible at global levels.

We recently designed a general framework capable of reconstructing evolutionary history directly from the structure of macromolecules. The framework enables global bottom-up or top-bottom approaches of genomic analysis and is supported by three fundamental premises: (1) that molecular structure is far more conserved than sequence and carries considerable phylogenetic signal, (2) that successfully implemented biological designs tend to be reused over and over again in nature, and (3) that there is a universal tendency towards molecular order. Bottom-up strategies unify phylogenetic analysis with structural biology using a Hennigian cladistic approach based on shared and derived features descriptive of common descent. Conversely, top-bottom strategies reveal global diversification using information embedded in entire genomic and proteomic complements. This enabled the charting of the protein world. In order to study protein diversity and evolution at a global scale, we counted the number of genes that could be assigned to particular protein architectures in genomes and used these measures of genomic demography to map the world of proteins and track architectural and organismal history at the proteome level.

Rooted phylogenies of proteomes and fold architectures were used to classify proteins, define structural transformations, determine general evolutionary trends in proteins structure, and study the evolution of metabolic and signaling networks. Phylogenetic tracings revealed patterns unique to multicellularity and inter-cellular signaling that could benefit the study of plant-microbial interactions.

Dr. Xinguang Zhu, a postdoctoral researcher in the UIUC Department of Plant Science, will speak on the topic of "Biological Cellular Metabolism Simulation in the Post Genomic Era."

Presentation slides: [Powerpoint] [PDF]

Abstract:

Cellular metabolism simulations are essential for not only improving our basic understanding about metabolisms, and improving our ability to engineering better metabolisms for defined purpose. The constraint based modeling, data based modeling, and kinetic modelings are the three main techniques used to model metabolisms. Each of these three different techniques needs different data or parameter sets, uses different algorithm, and can provide information about different aspects of metabolism, and inevitably has shortcomings of its own.

Photosynthesis, inarguably one of the most important biological metabolism on earth, is a exceptional model metabolism system in that it is well studied, with many measurable signals probing different sections of this process. Yet at the same time, it provides many opportunities for system biology research, e.g. the system properties nearly not studied, the interaction of this process with other metabolisms in leaf not well characterized, and the target photosynthesis gene to manipulate for high crop yield not identified. There are many challenges ahead that need to be solved before we can accurately simulate photosynthesis performance under different time scale and different environmental conditions, and pinpoint targets for improving photosynthesis efficiency.

Biography:

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.

Dr. Gary J. Olsen, Professor of Microbiology, will speak to the seminar on March 11. Dr. Olsen's research involves functions, evolutionary histories and structures of genes and proteins.

Title: A study of statistical methods for function  prediction of protein motifs

Speaker: Tao Tao, Department of Computer Science, UIUC Place: Siebel Center 3405 Time: 11am-12noon

Abstract:

Automatic discovery of new protein motifs (i.e.,  amino acid patterns) is one of the major challenges in  bioinformatics. Several algorithms have been proposed that  can extract statistically significant motif patterns from  any set of protein sequences. With these methods, one can  generate a large set of candidate motifs that may be  biologically meaningful. In this paper, we study several  statistical methods to automatically predict the functions  of these candidate motifs, including a popularity method, a  mutual information method, and statistical translation  models. These methods capture, from different perspectives,  the correlations between the matched motifs of a protein and  its assigned Gene Ontology(GO) terms, which characterize the  function of the protein. We evaluate these different methods  using the known motifs in the Interpro database. Each method  is used to rank candidate terms for each motif. We, then,  use mean reciprocal rank(MRR) to evaluate the performance.  The results show that in general, all these methods perform  well, suggesting that they can all be useful for predicting  an unknown motif s function. Among all the methods tested, a  statistical translation model with popularity prior performs  the best.

Bio:

Tao Tao is a Ph.D. candidate in the Computer Science  Department at University of Illinois at Urbana-Champaign.  He has been working on information retrieval models,  text/data mining with applications to bioinformatics.

Sameer Varma, Ph.D. candidate in Biophysics and Computational Biology at UIUC, will speak to the seminar on April 1. Varma has research interests in grid-based scientific computing and theoretical calculations of pH-dependent proton dissociation probabilities (pKa).

Biography:
Sameer Varma received his BS and MS degrees in physics from the Indian Institute of Technology in 1999. He is currently pursuing his PhD entitled "Protonation and Membrane Protein Function" in Biophysics and Computational Biology under the guidance of Prof. E. Jakobsson.

Title:
Structure-Function Relationships in Membrane Proteins.

Abstract:
Science may be defined as an attempt to find ever more fundamental laws and to reconstruct the long chains of causes from these foundations up to the full range of natural events. In adding its links to the chain, each scientific discipline adopts a certain phenomenon to work on at a given level of complexity and develops "fundamental" rules that can be considered a satisfactory explanation of what is seen at that level. These fundamental rules, however, to another scientific discipline might be considered a yet another complex phenomenon needing explanation. This has ever since been the trend in scientific endeavor and the science of biological molecules is no exception. This trend in the science of biological molecules would be presented in the context of our investigations of the structure-function relationships of two membrane proteins a membrane-spanning pore forming protein (OmpF) found in the outer-membrane of some single-celled organisms, and a signal-transducing membrane docking protein (C2 domain of cPLA₂) found in most multi-cellular organisms. Such investigations of biological molecules have been made possible not only due to pioneering work in the field of structural biology which have provided us with atomic resolution data for biological molecules, but also due to the worldwide efforts in the field of computer science and engineering that have today provided us with enormous computer power to solve complex mathematical equations.

Christine Elsik, assistant professor of Animal Sciences at Texas A&M and director of that university's Animal Bioinformatics and Computational Laboratory (http://racerx00.tamu.edu), will speak to the Seminar on April 8.

Title: Protein Clustering to Assemble Families of Homeomorphic Proteins

Presentation slides: [Powerpoint]

Abstract: With the capacity for genome projects to generate thousands of protein sequences per year, we need an effective automated method to group these sequences into families of proteins with identical domain organization (homeomorphic). Assembling families allows us to perform phylogenetic analysis to identify orthologs. Many proteins are modular, consisting of more than one structural domain. The combination of domains in a protein determines its function. Current protein clustering algorithms do not effectively separate multidomain proteins that share a common domain, but differ in overall domain organization. One reason for the difficulty is that we do not know the domain organization of many novel proteins, but rely merely on amino acid sequence similarity to classify proteins. I will compare hierarchical protein clustering algorithms and present a hidden Markov model for identifying protein domains using only sequence information.

Speaker: Xin He is a graduate student in the Department of Computer Science. He is working with Prof. Bruce Schatz and Chengxiang Zhai on bioinformatics and computational biology.

Title: Identifying Conserved Gene Clusters in the Presence of Homology Families

Abstract: This talk is based on a paper that will published in Journal of Computational Biology, Special Issue of RECOMB2004. The study of conserved gene clusters is important for understanding the forces behind genome organization and evolution, as well as the function of individual genes or gene groups. In this talk, I will present a model and algorithm for identifying conserved gene clusters from pairwise genome comparison. This generalizes a recent model called “gene teams”. A gene team is a set of genes that appear homologously in two or more species, possibly in a different order yet with the distance of adjacent genes in the team for each chromosome always no more than a certain threshold. We remove the constraint in the original model that each gene must have a unique occurrence in each chromosome, and thus allow the analysis on complex prokaryotic or eukaryotic genomes with extensive paralogs. Our algorithm analyzes a pair of chromosomes in O(mn) time and uses O(m + n) space, where m and n are the number of genes in the respective chromosomes. We demonstrate the utility of our methods by studying two bacterial genomes, E. coli K-12 and B. subtilis. Many of the teams identified by our algorithm correlate with documented E. coli operons, while several others match predicted operons, previously suggested by computational techniques. Our implementation and data are publicly available at http://euler.slu.edu/~goldwasser/homologyteams/.

Related Link:
https://netfiles.uiuc.edu/xinhe2/www/JCB.pdf
https://netfiles.uiuc.edu/xinhe2/www/RECOMB2004.pdf

Dr. Xinghua Lu is an Assistant Professor at the Dept of Biostatistics, Bioinformatics and Epidemiology of Medical University of South Carolina.  He was trained in Pharmacology and work in the field of bioinformatics after NLM sponsored postdoctoral training in Biomedical Informatics.  His research interests concentrates on applying latent variable models to simulate biological signaling system and text mining.

Abstract:  Knowledge related to proteins serves as a corner stone of modern biomedical knowledge.   In this talk, I will discuss applying statistical approaches to identify the latent semantic topics from a corpus of MEDLINE titles and abstracts describe biological aspects of proteins. A Bayesian model selection approach was employed to determine the optimal number of topics to represent the corpus. The identified latent topics were semantically coherent and majority of them reflected the biological concepts.  Furthermore, the latent semantic topics were mapped to the controlled vocabulary of the Gene Ontology.