BioLearn 2010

Evolutionary Computation and Machine Learning in Bioinformatics

9:30am to 12:30pm


December 2 (Thu)

Call for papers (txt)
Call for papers (html)

Technical Program

1. Rezarta Islamaj Doğan
2. Louis Licamele
3. Kenneth De Jong
4. Amarda Shehu
5. Uday Kamath

1. Rezarta Islamaj Doğan

Title: Machine Learning and Bioinformatics: An essential relationship with mutual benefits

Abstract: The field of machine learning started with the goal of building systems that can adapt to their environment and learn from their experiences. Now, after decades of intensive research, machine learning offers a powerful toolset of techniques that are playing a critical role in many industrial and scientific disciplines. Machine learning techniques are commonly invoked in various disciplines to solve the arising problems in a supervised, unsupervised, or reinforcement learning setting. Bioinformatics, on the other hand, started with the goal of building systems that can advance the understanding of biological processes and biological systems. Bioinformatics involves development of specialized databases and computational and statistical techniques to solve practical problems that arise from the analysis and management of all kinds of biological data.This talk aims at building a bridge between machine learning techniques such as dynamic programming, Bayes rule, and support vector machines, and their practical bioinformatics applications, such as genome annotation, gene expression analysis, and protein structure prediction. I will present the machine learning methods emphasizing both their general utility and their use for the bioinformatics applications. I will also address critical assessment, validation and evaluation methods that are used in bioinformatics, and survey recent evaluation initiatives as time permits.

Biosketch: Dr. Rezarta Islamaj Doğan is a Research Fellow in the Computational Biology Branch at the National Center for Biotechnology Information (NCBI / NLM / NIH). She joined NCBI after earning a PhD in Computer Science at the University of Maryland at College Park, where she was a member of the LINQS Machine Learning Research Group. Her research interests encompass machine learning and data mining approaches for identifying useful information in biomedical databases. She is interested in understanding user needs and their search habits in PubMed for improving biomedical information retrieval. She is also interested in discovering and building domain appropriate features in order to model the biomedical information for accurate classification and prediction.

2. Louis Licamele

Title: Machine Learning Applications in Bioinformatics

Abstract: Advancements in high throughput technology in the life sciences are continuing to be made and the amount of data being generated is only increasing. Scientists cannot hope to keep up without help. In order for knowledge to grow at a similar rate as the data being generated we must turn to informatics and machine learning for help. Automated pipelines, especially those that inform us of the most likely novel insights, are the key tools to advancing drug discovery and personalized medicine. In this talk several real life applications will be discussed, including the prediction of a novel mechanism of schizophrenia from gene expression signatures of antipsychotic drugs.

Biosketch: Louis Licamele is the Head of Informatics at Vanda Pharmaceuticals and oversees the bioinformatics, clinical data management, statistics, and IT functions. He helped set up the core informatics services in 2003. Prior to being the Head of Informatics, he served as Associate Director, Computational Research and Development. Mr. Licamele has conducted research in the areas of data mining and biomedical informatics, including the development of analytical methods for microarrays. He holds a B.S. in both Biology and Computer Science from Georgetown University and is currently completing his doctoral thesis in Computer Science from the University of Maryland on computational methods for drug discovery.

3. Kenneth De Jong

Title: Evolutionary Computation: A Unified Perspective

Abstract: The field of Evolutionary Computation has experienced tremendous growth over the past 15 years, resulting in a wide variety of evolutionary algorithms and applications. The result poses an interesting dilemma for many practitioners in the sense that, with such a wide variety of algorithms and approaches, it is often hard to see the relationships between them, assess strengths and weaknesses, and make good choices for new application areas. This presentation is intended to give an overview of EC via a general framework that can help compare and contrast approaches, encourages crossbreeding, and facilitates intelligent design choices. The use of this framework is then illustrated by showing how traditional EAs can be compared and contrasted with it, and how new EAs can be effectively designed using it.

Biosketch: Dr. De Jong is a Professor of Computer Science and Associate Director of the Krasnow Institute at George Mason University. Dr. De Jong's research interests include evolutionary computation, adaptive systems and machine learning. He is an active member of the Evolutionary Computation research community with a variety of papers, Ph.D. students, and presentations in this area. He is also responsible for many of the workshops and conferences on Evolutionary Algorithms. He is the founding editor-in-chief of the journal Evolutionary Computation (MIT Press), and a member of the board of ACM SIGEVO. He is the recipient of an IEEE Pioneer award in the field of Evolutionary Computation and a lifetime achievement award from the Evolutionary Programming Society.

4. Amarda Shehu

Title: Combining Evolutionary Algorithms and Machine Learning in Bioinformatics

Abstract: This talk will provide an overview of recent computational methods that combine evolutionary algorithms with machine learning techniques for annotation of biological sequences and structures. Many problems of interest in bioinformatics necessitate annotation of biological molecules with structural or functional information. For instance, determining that a protein is not crystallizable from its sequence can save futile attempts to determine the structure through X-ray crystallography. Yet other problems involve predicting splice sites and regulatory regions in DNA molecules. Annotation often involves extracting descriptive features, a task that has commonly been the domain of problem experts or a combination of painstaking efforts and luck. Recent research advocates employing evolutionary algorithms for the automatic discovery of meaningful biological features. This talk will provide an overview of a general emerging framework that combines evolutionary algorithms with machine learning techniques and showcase several successful applications of this framework in bioinformatics.

Bio: Dr. Amarda Shehu is an Assistant Professor in the Department of Computer Science at George Mason University. She holds affiliated appointments in the Department of Bioinformatics and Computational Biology and the Bioengineering Program at George Mason University. She received her Ph.D. in Computer Science from Rice University in Houston, TX in 2008, where she was also an NIH fellow of the Nanobiology Training Program of the Gulf Coast Consortia. Her research in computational biophysics and bioinformatics focuses on advancing understanding of the sequence-structure-function relationship in biological molecules. Her work combines probabilistic search frameworks with the theory of statistical mechanics and evolutionary algorithms with machine learning.

5. Uday Kamath

Title: Annotation of DNA Sequences with Evolutionary Algorithms and Support Vector Machines

Abstract: Hybrid methodologies that combine supervised learning techniques and evolutionary algorithms are often used in the Bioinformatics domain to annotate or classify biological sequences. Support Vector Machines (SVMs) are the most effective and popular supervised learning technique in binary classification problems. Effective application of SVMs on biological sequences, however, depends on extracting descriptive features from the biological sequences. In this talk I will showcase some of our recent research on detection of hypersensitive and splice sites in DNA sequences. Our methodologies combine evolutionary techniques like genetic algorithms and genetic programming with SVMs. Our applications show that this combination extracts meaningful and novel biological features that significantly improve SVM classification of DNA sequences in the context of recognition of hypersensitive and splice sites .

Bio: Uday Kamath is a Ph.D Student in the Department of Computer Science at George Mason University. He also works as Technical Architect in the Financial Analytics Department of Norkom Technologies. He has a Masters in Computer Science from the University of North Carolina at Charlotte. His main research focus is the combination of evolutionary techniques with machine learning and their application on complex problems in Bioinformatics and financial crime.