Bayesian classification of protein structure

Abstract
The application of an Autoclass III machine-learning algorithm for heuristic Bayesian classification to independently confirm and provide new information about structural classes of proteins is discussed. Bayesian classification and Autoclass III systems are reviewed. The decisions concerning and data representations and where to begin searching the space of potential classifications are discussed. Results of the application, in terms of classification patterns, relationship to traditional classes, higher-order structure, and class composition, are presented.<>