The astonishing, exponentially increasing rates of genome sequencing has led to one of the most significant challenges for the biological and computational sciences in the 21st century: assigning the likely functions of the encoded proteins. Enzymes represent a particular challenge, and a critical one, because the universe of enzymes is likely to contain many novel functions that may be useful for synthetic biology, or as drug targets. Current approaches to protein annotation are largely based on bioinformatics. At the simplest level, this annotation involves transferring the annotations of characterized enzymes to related sequences. In practice, however, there is no simple, sequence based criterion for transferring annotations, and bioinformatics alone cannot propose new enzymatic functions. Structure-based computational methods have the potential to address these limitations, by identifying potential substrates of enzymes, as we and others have shown. One successful approach has used in silico 'docking' methods, more commonly applied in structure-based drug design, to identify possible metabolite substrates. A major limitation of this approach is that it only considers substrate binding, and does not directly assess the potential of the enzyme to catalyze a particular reaction using a particular substrate. That is, substrate binding affinity is necessary but not sufficient to assign function. A reaction profile is ultimately what is needed for a more complete quantitative description of function. To address this rather fundamental limitation, they propose to use quantum mechanical methods to explicitly compute transition state barriers that govern the rates of catalysis. Although quantum mechanical, and mixed quantum/classical (QM/MM), methods have been used extensively to investigate enzymatic reactions, the focus has been primarily on elucidating complex reaction mechanisms. Here, the key catalytic steps are known, and they use these methods quantify substrate specificity. That is, we bring the power of quantum mechanics to bear on the problem of annotating enzyme function, which is a novel approach. Although it has been clear to us at the Jacobson group for some time that enzyme specificity may be encoded in transition states, rather than simply substrate recognition, the main limitation has always been computational expense. Using a hierarchy of different methods, they can reduce the list of plausible substrates of an enzyme to a small number in most cases, but even identifying the transition states for a dozen plausible substrates requires significant computational effort, beyond what is practical using standard QM/MM methods. For this project, they have chosen two enzyme superfamilies which they have used as 'model systems' for functional assignment. The enolase superfamily is a large group of {alpha}-{beta} barrel enzymes with highly diverse substrates and chemical transformations. Despite decades of work, over a third of the superfamily remains unassigned, which means that the remaining cases are by definition difficult to assign. They have focused on acid sugar dehydratases, and have considerable expertise on the matter. They are also interested in the isoprenoid synthase superfamily, which is of central interest to the synthetic biology community, because these enzymes are used by nature to create complex rare natural products of medicinal value. the most notable example of this is the artemisinin, an antimalarial compound that is found in trace amounts in the wormwod root. From the standpoint of enzyme function assignment, these enzymes are intriguing because they use a small number of chemically simple substrates to generate, potentially, tens of thousands of different products. Hence, substrate binding specificity is only a small part of the challenge
the key is determining how the enzyme directs the carbocation chemistry to specific products. These more complex modeling approaches clearly require quantum mechanical methods.