Nature has generated an impressing diversity of biocatalysts varying in sequence and biochemical properties. This biodiversity is being exploited by screening of cultivable and uncultivable organisms for new enzymes with improved properties. Protein engineering is a powerful tool to further improve activity, substrate specificity, and regio- or stereoselectivity of recombinant enzymes. Understanding the relationship between sequence, structure, and function of enzymes would greatly help screening and engineering; conversely, large-scale experimental data provide an invaluable knowledge base for developing molecular models of enzymes. To study the relationship of sequence, structure, and function of enzymes, we have combined two approaches: the systematic analysis of a comprehensive family-specific database which integrates annotated sequence and structure information, and molecular modeling of enzyme-substrate interactions by docking and molecular dynamics simulations.

The Lipase Engineering Database has been established as a tool for systematic analysis of more than 1500 a/b hydrolases and to identify structurally and functionally relevant modules¹. The existence of highly conserved structural modules allowed to reliably model 80% of all epoxide hydrolases in the database², even at a level of less than 20% sequence identity. In addition, functional modules were identified which mediate substrate specificity: while most hydrolases with a GGGX motif accept esters of tertiary alcohols, GX-types are inactive towards this substrate class³. For GX-types, the concept of functional modules was further extended, and residues were identified which are essential for catalysis but are not in contact with the substrate.

By combining docking and molecular dynamics simulations, a general molecular model of enantioselectivity of a/b hydrolases was established⁴. The model is general and predictive: a broad variety of substrates were ranked by selectivity and improved mutants in the substrate binding site were predicted^5,6. Recently, molecular dynamics simulations were successfully applied to reproduce long-range effects of mutations in a metallo-b-lactamase⁷.