ABSTRACT
The 'Protein Folding Problem' is probably the greatest remaining challenge in structural molecular biology. There are three main categories of protein structure prediction methods: comparative modelling methods, fold recognition methods and ab initio methods. In comparative modelling, the structure of the new protein is predicted by comparing its sequence with the sequences of proteins of known structure. Ab initio methods are so-called because they use only the information in the target sequence itself. There are two branches of ab initio prediction: knowledge-based methods; and simulation methods. The chapter discusses the prediction of protein secondary structure. There are two aspects of protein structure that are exploited in predicting secondary structure from just a single protein sequence: the intrinsic secondary structure propensities of amino acids, and the hydropathic nature of amino acids. The most effective method of secondary structure prediction is neural networks used to analyze substitution patterns in multiple sequence alignments.
