A new automated method for modeling non-covalent macromolecular interactions

Molecular recognition and non-covalent macromolecular associations are important events in Chemistry and Biology. The prediction of the structure of complexes formed within macromolecular systems is a valuable step in the understanding of chemical pathways and of mechanisms operating in living organisms. A new effective and automated method is reported for predicting the most likely mode of binding of two proteins, using the three-dimensional structures of the interacting molecules. The program parameterized and tested on a series of protein complexes, well documented by crystallographic data. For thirteen out of fifteen tested cases, it is shown to successfully predict the correct or an approximate binding geometry. The value and limitations of the method for the study of strong and weak macromolecular interactions, not yet revealed by experimental techniques, are discussed. Most of the time consuming tasks rely on very fast Boolean type (binary, such as OR, AND, etc,) operations, which makes it feasible to run on inexpensive and widely accessible personal computers.