Conference Program

Tutorial Track 4

Title:  Computational intelligence Techniques for Protein Analysis
Presenter: Dr. Gwenn Volkert, Kent State University
Abstract:  There is growing interest in the application of computational intelligence (CI) methods to problems in bioinformatics.  Specifically, as the amount of data available increases there is an appreciation that many of the problems in bioinformatics need to be addressed in more informed and intelligent ways. For instance, while protein-folding prediction from first principles is feasible for short protein sequences (~20 amino acids), fold prediction for longer, biologically realistic, sequences (200 or 300 amino acids and more) rapidly becomes intractable.

Computational intelligence (CI) is an area of computer science that specializes in dealing with problems typically considered intractable through the use of heuristics and probabilistic approaches.  CI approaches are ideal for problems where there is not a precise definition for an "absolutely provably correct or best" answer such as would be obtainable through the use of operations research type analysis (i.e. "hard" computing).  Rather, CI is ideal where the requirement is for an answer that is better than one currently known, or which is acceptable within certain defined constraints (i.e. "soft" computing). Given that many problems in bioinformatics lack absolutely correct or best answers there are many appropriate applications for CI methods in bioinformatics.

Despite the apparent suitability of, and in fact successful application of CI techniques to bioinformatics problems, there is actually very little published on the details of such applications when compared to the nearly exponentially increasing amount of papers published in bioinformatics annually. The objective of this tutorial is to introduce bioinformatics researchers to two particular CI techniques, neural networks and evolutionary algorithms. Current bioinformatics textbooks and much of the published literature typically do not provide detailed explanations of these techniques, leaving the reader to intuit that the author is sufficiently knowledgeable in the application of these techniques as a matter of blind trust. The tutorial will introduce the basic algorithmic properties of neural networks and evolutionary algorithms and then demonstrate their use through the presentation of examples taken from the bioinformatics literature.

The overall goal of the tutorial is to provide participants with enough introductory knowledge and insight to enable successful use of and ability to assess the potential for the application of neural networks or evolutionary algorithms to a variety of bioinformatics application areas. The specific examples presented will focus on protein data analysis but parallels to other types of bioinformatics data will be made. Additionally, "discoveries" facilitated by the use of these techniques will be highlighted to demonstrate the value of applying CI methods to bioinformatics problems.

Back to Schedule