Protein fold class prediction is a new field for statistical classificarion and regression

摘要

Protein structure classification and prediction is introduced and elaborated for the application of standard and new statistical classification, discrimination and regression methods. With the sequence to structure to function paradigm in the background, methods of secondary and tertiary structure prediction will be reviewed and super-secondary classes and of fold classes will defined. We apply two branches of statistical classification - methods based on posterior probabilities and methods based on class conditional probabilities - and we will explore the role of artificial neural networks for the protein structure prediction. The procedures will be applied to a set of 268 previously described protein sequences for their statistical performance in the prediction of the four supersecondary classes and also in the prediction of 42 fold structure classes.