Files
Abstract
Several AI techniques are applied in two scientific task domains. Genetic Programming (GP) is used to evolve a set of functions to approximate the static dielectric constant of water and several different binary classification algorithms are compared in their ability to distinguish translation start sites on two different prokaryotic genomes. GP performs very well as compared with standard statistical approaches to approximating the dielectric constant, and is a very powerful new tool that can be used for regression analysis in this and related domains. Translation start site prediction remains an open problem in bioinformatics, and several computational models for translation start site prediction have been created before. Support vector machines, decision trees, nave bayes, artificial neural networks, and XCS are all compared in their ability to locate translation start sites. XCS has never been used for this task and performs as well as the other aforementioned techniques, making the technique a viable new candidate for generating predictive models for this and other computational biological problems.