Bioinformatics Research Group
Bioinformatics research at the University of New Orleans benefits from excellent ties with the Research Institute for Children (RIC) and from critical support from the Board of Regents Enhancement Program. RIC is a research facility administered jointly by LSUHSC (Louisiana State University Health Sciences Center) and Children's Hospital, New Orleans. The primary faculty for the Bioinformatics group: Dr. Winters-Hilt, Dr. Summa, Dr. Zhu, and Dr. Taylor, each have ties with the Research Institute. Dr. Winters-Hilt and Dr. Summa operate the primary wet-lab component of their research at facilities located at RIC. The primary computational facilities for the Bioinformatics Group are at the Computer Science Department's Interdisciplinary Center for Bioinformatics (ICfB), which is located on the 2nd floor of the UNO Center for Energy Resource Management (CERM), and consists of several rooms and meeting spaces. The CERM computational facilities include a large, state-of-the-art computational cluster, consisting of 27 dual AMD64 machines with 4G+ RAM and Gig-Ethernet connectivity. A new Linux cluster consisting of 21 1U rack units (possibly up to 38 units), each containing dual-CPU dual-core 2.6GHz AMD Opteron processors and 8G RAM, is currently in the process of being built. A SunFire V880 with four 16G RAM UltraSPARC CPUs, and six 73GB SCSI Drives, is currently being incorporated into the computational cluster to provide an asymmetric distributed processing platform that is optimal for many AI and Machine Learning algorithms, and for real-time signal processing. Also located at CERM is a specialized wet-lab operated by Dr. Winters-Hilt for pattern recognition informed nanopore detector and biophysics experiments.
UNO's Bioinformatics facilities are being rapidly expanded via two recent LA Board of Regents Enhancement Grants: (1) $364,398 (incl. Match funds) to develop an Interdisciplinary Center for Bioinformatics, PI's: Stephen Winters-Hilt and Mahdi Abdelguerfi; and (2) $287,652 (incl. Match funds) to develop a Multidisciplinary Program in Molecular Simulation, Visualization, and Engineering, PI's: Christopher Summa, Stephen Winters-Hilt, Steven Rick, and Mahdi Abdelguerfi. The other CS Departmental facilities offer dozens more operational nodes for cluster computations (depending on the RAM cutoff imposed), providing an approximate doubling of the computational capabilities of the bioinformatics-dedicated cluster facilites. In addition to this, there are recently added University-based facilities for use by UNO co-PI's on the LONI 'Supercomputer' project (UNO-based co-PI's: Vassil Rousssev, Mahdi Abdelguerfi, Stephen Winters-Hilt) and the PETAshare 'Superstorage' project (UNO-based co-PI's: Stephen Winters-Hilt and Mahdi Abdelguerfi). UNO-based bioinformatics researchers have priority access to the aforementioned 'super' facilities.
Student researchers are making significant contributions to many of the research efforts described in what follows. There are currently eighteen graduate students and one undergraduate concentrating on bioinformatics. Bioinformatics is a well-funded, rapidly growing area. All students in the bioinformatics concentration currently have tuition and stipend covered from grant funds.
Dr. Stephen Winters-Hilt
I'm interested in biophysics and bioinformatics problems that substantially benefit from application of advanced, machine learning (artificial intelligence) based statistical analysis tools. This allows two complimentary pursuits, one to develop novel AI-based pattern recognition tools, the other to explore the realm of the biophysics or bioinformatics application. The bioinformatics and computational biophysics/biochemistry efforts involve a large and growing number of researchers -- the Winters-Hilt Group currently has fourteen student researchers (graduate and undergraduate), two lab technicians, and two postdocs.
The main application of interest at this time focuses on the device physics of the alpha hemolysin nanopore detector, and application of that detector to a variety of single-molecule biophysical/biochemistry binding studies and conformational change studies. Antibody-antigen binding studies, for example, are the focus of a three year NIH K-22 grant (PI), and DNA conformational studies are the focus of a 2.5 year NIH R-21 grant (co-PI).
The kinetically modulated channel current signal measured in the nanopore experiments is rich with information, and yet simply and cheaply obtained -- tens of thousands of dollars for a nanopore detector versus millions for equipment used in other single-molecule experiments. To get at nanopore detector information effectively, however, requires a variety of statistical learning tools, including Hidden Markov Models,(HMMs), Support Vector Machines (SVMs), and Finite State Automata (FSAs). Bioinformatics/AI-tools for analysis of stochastic sequential data is a strong area of interest in general, with the focus here on channel current analysis (to enable nanopore detection) and genome-wide gene structure prediction. The latter application takes on particular novelty when coupled with nanopore detector based verification of any purported transcription factor binding sites. My overall research interests/applications include Machine Learning (AI) based signal analysis and pattern recognition; nanopore detector device physics & biotechnology applications; single molecule biophysics & biochemistry; genome-wide gene structure identification (bioinformatics in general); on-line learning; and financial analysis.
- Churbanov, A. and S. Winters-Hilt, "Implementing EM and Viterbi algorithms for Hidden Markov Model in linear memory," Accepted by BMC Bioinformatics (2008).
- Winters-Hilt S., "Nanopore Cheminformatics based Studies of Individual Molecular Interactions." Ch. 13 in Yanqing Zhang and Jagath C. Rajapakse, editors, Machine Learning in Bioinformatics, John Wiley & Sons, 2007.
- Winters-Hilt, S., Channel current cheminformatics and bioengineering methods for immunological screening, single-molecule analysis, and single molecular-interaction analysis . PATENT, UNO filing, 2005, with priority established via four Provisional PATENT filings in 2004.
- Winters-Hilt S., W. Vercoutere, V. S. DeGuzman, D. Deamer, M. Akeson, and D. Haussler, "Highly Accurate Classification of Watson-Crick Base-Pairs on Termini of Single DNA Molecules," Biophys. J. Vol. 84, pg 967, 2003.
Dr. Christopher M. Summa
I am particularly interested in methods development and applications in the fields of bioinformatics and computational structural biology. The function of biological molecules is dictated in part by their 3-dimensional structure and dynamics, and we develop and use computational tools to study these properties. We use both simulation techniques on individual molecules and data-mining and statistical techniques to test theories about protein structure and function.
- Theoretical and Experimental studies of protein-protein interactions
- Protein Structure Modeling and Prediction
- Protein Design and Engineering
These studies involve:
Summa CM and Levitt M
Near Native Structure Refinement Using in vacuo Energy MinimizationProceedings of the National Academy of Sciences USA 2007 (104) 3177-3182
Summa CM, DeGrado WF, and Levitt M
An Atomic Environment Potential for use in Protein Structure PredictionJournal of Molecular Biology 2005 (352) 986-1001
Slovic AM, Summa CM, Lear JD, DeGrado WF,
Computational Design of a Water-Soluble Analog of Phospholamban,Protein Science 2003 (12) 337-348
Dr. Dongxiao Zhu (http://cs.uno.edu/~dzhu)
I'm interested in both methodology research and it's applications to solve fundamental biological problems using ever-accumulating genome-wide data. My research mainly lies in the interface between contemporary molecular biology, computational science and statistics.
One of my current research concerns reverse engineering methodology to infer biological pathways and networks from high throughput data. Biological pathways/networks serve as a primary means to regulate cell growth, differentiation and apoptosis. Unfortunately, it is difficult to obtain data that directly reveal network topology and so reverse engineering is a viable method to uncover the underlying bio-complexity.
Another research interest concerns developing and tailoring data mining and pattern recognition methods to analyze genome-wide data. More specifically, I am attempting to address some statistical/mathematical issues arising from the large p, small n paradigm. For example, using a constrained learning and/or shrinkage method, where constraints are inspired from real-world biology prior knowledge or network topology.
- Zhu, D., Li, Y. and Li, H. 2007. Multivariate Correlation Estimator for Inferring Functional Relationships from Genome-Wide Data. Bioinformatics, 23(17), 2298-2305.
- Zhu, D. and Hero, A.O. 2007. Bayesian Hierarchical Model for Large Covariance Matrix Estimation. J. Comput. Biol., 14(10), 1311-1326.
- Zhu, D., Hero, A.O., Cheng, H., Khanna, R. and Swaroop, A. 2005. Network Constrained Clustering for Gene Microarray Data. Bioinformatics, 21(21), 4014-4020.
- Zhu, D., Hero, A.O., Qin, Z.S. and Swaroop, A. 2005. High Throughput Screening Co-expressed Gene Pairs with Controlled FDR and MAS. J. Comput. Biol. 12(7), 1027-1043.
Dr. Christopher M. Taylor
My research interests involve the development of algorithms and computational techniques to analyze and interpret biological data on a genomic scale. A main focus of my recent work is designing algorithms for the application of genome-tiling microarrays to human cancer research. Effective analysis of these experiments requires development and application of discrete algorithms in tandem with statistical techniques. These methods allow us to generate high resolution profiles of DNA replication timing and identify potential sites of DNA replication origin in human cancer cells. This work has been integrated with concurrent research around the world as part of the NIH directed ENCODE consortium and is being expanded to address other outstanding issues in human cancer research.
The development of high-throughput sequencing technologies promises to move us ever closer to the much sought-after $1000 genome. These new sequencing methods are currently enabling biologists to perform studies that would not be feasible with traditional capillary sequencing. There is a great need for custom algorithm development to efficiently analyze this new sequencing data and to help these new technologies achieve their full potential. I am collaborating directly with molecular biologists who perform these experiments to develop analysis algorithms. This work has strong implications not only for basic scientific research, but also in the public health sector which will begin to use these techniques in the future for personal health care as the cost of sequencing the human genome continues to decrease.
- Karnani N, Taylor C, Dutta A. Microarray Analysis of DNA Replication Timing. To appear in "Microarray Analysis of the Physical Genome". Methods in Molecular Biology. Humana Press, 2008.
- Encode Project Consortium. Identification and Analysis of Functional Elements in 1% of the Human Genome by the Encode Pilot Project. Nature. 2007 Jun 14;447(7146):799-816.
- Karnani N, Taylor C, Malhotra M, Dutta A. Pan-S Replication Patterns and Chromosomal Domains Defined by Genome-Tiling Arrays of Encode Genomic Areas. Genome Research. 2007 Jun;17(6):865-76.
- Encode Project Consortium. The Encode (ENCyclopedia Of DNA Elements) Project. Science. 2004 Oct 22;306(5696):636-40.