Seminar Series

Computer Science Seminar Series - Spring 2007
Date/Time	Speaker	Title	Institution	Location
3/14/07 2-3PM	Heng Huang	Marker Gene Selection and Gene Regulatory Elements Identification in Microarray Data Analysis	University of Texas at Arlington	DION 101
4/25/07 2-3PM	Li Shen	Parametric Surface Modeling and Morphometric Pattern Analysis	UMassD	DION 101
4/27/07 3-5PM		Graduate Students - Research Projects	UMassD	DION 101
5/2/07 2-3PM		Graduate Students - Research Projects	UMassD
5/4/07 3-5PM		Graduate Students - Research Projects	UMassD	DION 101
5/9/07 2-3PM		Graduate Students - Research Projects	UMassD	DION 101
5/11/07 3-5PM		Graduate Students - Research Projects	UMassD	DION 101

3/14/07

Title: Marker Gene Selection and Gene Regulatory Elements Identification in Microarray Data Analysis

Abstract
Microarray gene expression techniques have recently made it possible to offer phenotype classification of many diseases. One problem in this analysis is that each sample is represented by quite a large number of genes, and many of them are insignificant or redundant to clarify the disease problem. The previous work has shown that selecting informative genes from microarray data can improve the accuracy of classification. We proposed to use a heuristic K-means+ based clustering method to group similar genes and select informative genes from them to avoid redundancy and extract biological information from them. Empirical results on both public data sets and on data sets in the MAGIC database (Dartmouth Medical School) have shown that our methods can successfully select good candidate marker genes that lead to better classification accuracy than other methods. Meanwhile we also presented a new test environment in which synthetic data is perturbed to simulate possible variations in gene expression values. The goal is for the generated data to have appropriate properties that match natural data, and that are appropriate for use in testing the sensitivity of feature selection algorithms and validating the robustness of selected marker genes.

On the other research work, we study the organization of protein binding sites (cis-regulatory elements) in sets of co-regulated genes. The identification of potential cis-regulatory elements in the upstream regions of genes is key to understanding the mechanisms that regulate gene expression. We presented the efficient algorithms -- beam-search enumerative algorithm for motif finding, which aimed at the discovery of cis-regulatory elements in the DNA sequences upstream of a related group of genes. This algorithm dramatically limits the search space of expanded sequences, converting the problem from one that is exponential in the length of motifs sought to one that is linear. Unlike sampling algorithms, our algorithms converge and are capable of finding statistically overrepresented motifs with a low failure rate.


Speaker Bio

Heng Huang received his Ph.D. in Computer Science from Dartmouth College in 2006, and his M.S. and B.S. from Electronics and Information School at Shanghai Jiao Tong University. He was working on bioinformatics and system biology as a postdoctoral researcher in the Center of Biological and Biomedical Computing at Biological Department, Dartmouth Medical School. His contribution has been in the areas of computer vision, biomedical image computing, bioinformatics, and scientific visualization. He is director of the BIOVIZION Lab (Biomedical Computing and Scientific Visualization Laboratory) in Computer Science and Engineering Department at University of Texas at Arlington.


Date TBD

Title: Genetic Networks: from analysis and complexity to biotechnology

Abstract

The concept of a ``genetic network'' refers to the complex network of interactions between genes and gene products in a cell. These regulatory networks have thus the task of controlling all aspects of life. Recently synthetic genetic networks were constructed and this initialized a. call to understand their computational power. With A. Ben Hur (2004) we considered a model of analog gene networks and proved that they are equivalent computationally to Turing machines. Unlike the neural networks case, these are robust with respect to system’s perturbation. While the GN equations are in general chaotic, we proposed a particular design principle that makes them non-chaotic and easy to fabricate. This work was described in PHYSICS NEWS UPDATE : January 2004.

With F. Roth (PhD thesis 2006) and R. Douglas (2007), we took a different approach to study the genetic network. We designed an artificial stem cell and let if develop in a physical like environment. We demonstrated how the stem cell developed into an artificial organism with sensors and motor cells, axons, and clear behavior. We also showed how different environment and different feedbacks during development may lead to crucial differences in the structure and the emerging behavior.

* *The interaction between gene activation and cellular activity is a critical aspect of emergent behavior. As a step towards unfolding this relation we focus on the circadian system. With T. Leise we raised the question of what causes jet-lag to occur, and how biological findings of “chaos” among the different tissues during jetlag and after unnatural shift work can be explained. We proposed an analytical model based on multistage nonlinear dynamics. The modeling predicts that jet lag tends to be most severe following an eastward change of 5-8 time zones due to prolonged desynchrony of the system. This desynchrony is partly due to differing reentrainment rates among components, but a much greater source of desynchrony is the antidromic reentrainment of some but not all components, triggered by the overshoot of the master pacemaker's phase in response to these advances. Based on the multistage system dynamics, we design a simple protocol that results in a more orderly transition that avoids antidromic reentrainment in all components, thereby reducing the reentrainment time from nearly two weeks to just a few days for the most difficult shifts. We compare the predicted behavior of damped versus robust oscillatory components in the system, as well as the effect of weak versus strong coupling from the master pacemaker to the peripheral components. (2006). The work was described in much media including Yahoo news, Forbes, Globes, NP Radio and more.

Computer Science Seminar Series - Fall 2006 Time: 3PM - 4PM
Date	Speaker	Title	Institution	Location
10/13/06	Gary Livingston	Machine Learning Applications in Bioinformatics: Inferring Subpopulation Differences in Variable Interactions	UMass, Lowell	DION 101
11/03/06	Michael Geiger	Embedded Systems Design:  A Computer Architect’s Perspective	UMassD	DION 101
12/01/06	Azer Bestavros	Typed Representation and Analysis of Network Flows for Scalable and Practical Interoperability Checks	Boston University	DION 101
12/08/06		Graduate Students - Research Projects	UMassD	DION 101
12/15/06		Graduate Students - Research Projects	UMassD

10/13/06

Title: Machine Learning Applications in Bioinformatics: Inferring Subpopulation Differences in Variable Interactions

Abstract
Bayesian network learning and other methods have been adapted to infer interactions among variables, for example, inferring gene interactions from gene expression microarray data. However, these methods do not infer differences in variable interactions between subpopulations (e.g., gene A’s expression is positively correlated with gene B’s expression in normal liver tissue, but gene A’s expression is negatively correlated with gene B’s expression in cancerous liver tissue). This talk will present (1) a Bayesian network-based method for inferring interactions among variables that differ between subpopulations, (2) results from evaluating the method using artificially generated data and yeast gene expression microarray data, and (3) results from applying the method to breast cancer promoter data from which new significant hypotheses about differences in gene interactions between two subtypes of breast cancer, ERBB2+ (poor prognosis) and Normal-like (good prognosis) have been made.

Speaker Bio
Dr. Livingston is an Assistant Professor in the University of Massachusetts Lowell Department of Computer Science. His research interests are in using domain knowledge to aid Bayesian network structure learning and in scientific applications of machine learning. Most recently, Dr. Livingston is focusing his research into using additional sources of biological information, such as protein-protein interactions and genomic information, to aid the inference of genetic pathways from leukemia, lymphoma, and chlorella virus gene expression microarray data.


11/3/06

Title: Embedded Systems Design:  A Computer Architect’s Perspective

Abstract
Computer architects have effectively exhausted the design space
of general-purpose processors.  The search for new research has led many
architects to explore embedded designs, which must deliver high
performance within a restrictive energy budget. These constraints force
designers to consider novel approaches to system design, particularly from
an architectural perspective.  This talk will demonstrate the need for
further exploration of embedded architectures. To illustrate the impact of
embedded design principles, the lecture will focus primarily on prior
research involving data cache architectures, but it will also discuss
potential avenues of computer architecture research in the embedded
domain.


Speaker Bio
Dr. Geiger is an Assistant Professor in the Electrical & Computer
Engineering and Computer & Information Science Departments at UMass
Dartmouth.  A recent graduate of the University of Michigan, his general
research interests include computer architecture and embedded system
design.  His publications thus far have focused on the design of cache
architectures to improve energy consumption.


12/1/06

Title: Typed Representation and Analysis of Network Flows for Scalable and Practical Interoperability Checks

Abstract
The heterogeneity and open nature of networked systems make analysis of compositions of components quite challenging, consequently making the design and implementation of robust network services largely inaccessible to average network architects and network application programmers. In this talk I will overview a novel type system and associated type spaces, which constitute accessible representations of the results and conclusions that are derivable using complex compositional theories. These representations allow a networking system architect or programmer to be exposed only to the inputs and output of compositional analysis without having to be familiar with the ins and outs of its internals. Toward this end, I will present the TRAFFIC (Typed Representation and Analysis of network Flows For Interoperability Checks) framework, a simple flow-composition and typing language with corresponding type system. Next, I will discuss and demonstrate the expressive power of a type space for TRAFFIC derived from the network calculus, which allows us to reason about and infer such properties as data arrival, transit, and loss rates in large composite network applications. The TRAFFIC compositional analysis framework will be put in action using a prototype implementation of a type checking and inference engine, which is available for demonstration purposes using a web interface.
This work was pursued in collaboration with Adam Bradley, Yarom Gabay, Assaf Kfoury, Likai Liu, and Ibrahim Matta.



Speaker Bio
Azer Bestavros obtained his SM in 1988 and his PhD in 1992, both in Computer Science from Harvard University. He is currently Professor and Chairman of Computer Science at Boston University. His research interests are in the general areas of networking and real-time systems. Some of his seminal works include his generalization of classical rate-monotonic analysis to accommodate probabilistic guarantees, his pioneering of the push model for Internet content distribution adopted years later by CDNs, and his characterization of Web traffic self-similarity and reference locality. With over 2,000 citations to his publications, CiteSeer ranks him in the top 250 of its 10,000 most-cited CS authors at all times. His research has been funded by government and industry grants totaling over $15M. He received distinguished service awards from both the ACM and the IEEE, and is a distinguished speaker of the IEEE.