KNOWLEDGE-BASED PROTEIN MODELLING AND DESIGN

Abstract

Knowledge-based modelling can be envisaged as a number of steps concerned with the establishment and use of rules to generate a model of a protein. One of the most powerful procedures in learning rules is comparison of related Structures either through alignment of sequences to identify conserved residues or superposition of three dimensionalstructures to identify conserved conformations or motifs. Thus the first step in a knowledge-based modelling procedure is the systematic comparison of families of topologically similar structures. This step will lead to the establishment of "equivalences" between the structures compared and to their clustering based on measures of similarity. The second step involves the projection of the results of the comparisons of three dimensional structures down onto the level of sequence. This step establishes rules relating sequence to structure. These can be expressed as consensus sequences - templates - for topologically equivalenced residues, or as key residues in canonical structures, which are then used to align the sequence of the protein of unknown tertiary Structure. The third step uses the rules established in the second step to generate a three-dimensional model.