As mentioned above, one of the biggest limitations of classical MD simulations is the inability to make and break covalent bonds during a simulation. A potential solution to this is to describe the system quantum mechanically instead of classically. This would seem to be an ideal situation since then electron densities are explicitly included in the calculation. However, such simulations are far too computationally expensive to be used for molecular dynamics simulations of proteins. An alternative approach, first proposed in 1976 by Warshel and Levitt (59), was to combine a quantum mechanical (QM) potential with a molecular mechanical (MM) potential to form a hybrid QM/MM potential. In this approach, the parts of the protein and substrate that are directly involved in the enzyme reaction are calculated using QM potential functions, and the remaining atoms are treated using a classical MM potential. The coupling of a QM and MM potential allows just the reaction center to be studied quantum mechanically while keeping the calculation complexity low by using a more approximate MM potential elsewhere. This partitioning of the system allows calculations on systems significantly larger than would be possible with pure QM approaches and at the same time enables calculations such as reactions to be studied for which classical MM potentials are not appropriate. A number of commonly used MD codes provide support for QM/MM simulations including AMBER and CHARMM. QM/MM simulations still remain relatively expensive, however, although recent advances in AMBER (60) are bringing the cost of QM/MM simulations, for systems with up to 100 QM atoms, to a point where the cost is approximately double what the standard classical simulation would cost. Thus, QM/MM approaches will form an important tool for looking at the actual mechanism of hydrolysis within a cellulase enzyme as it degrades cellulose fibers.