Chemistry Department
Room 139 Smith Hall
207 Pleasant St SE
MN 55455
yorkx009@umn.edu

The specific aims of this work are to develop new multi-level quantum models for catalytic RNA systems, and to apply them to the metal-ion catalyzed phosphate hydrolysis reaction in the hammerhead ribozyme [21, 4, 15, 11] and related small models systems. These systems pose special challenges for computational models since they are complex, highly ionic systems, and involve phosphate groups that are more difficult to treat than first-row atoms. Consequently, methods need to be designed that can handle large solvated systems with quantum methods that are sufficiently accurate. The insights gained from the applications will have broad implications into other catalytic RNA systems.

In the first stage of the NIH-supported research, long-time classical molecular simulations are to be performed of the hammerhead ribozyme in several solution environments including: low, medium and high salt solutions, bound divalent metals at fractionally occupied crystallographic positions. In addition, crystal simulations of the native, metal-ion bound, catalytically trapped intermediate [11, 12] and cleavage product [10] will be performed in order to 1) make direct comparison with experiment to assess the quality of the force field and simulation protocol, and 2) to isolate the effects of solvation compared with the experimental crystalline environment. The simulations are performed using an academic version CHARMM c27 [1, 6].

These are large-scale simulations ranging in size from 23,000-55,000 atoms, and require a fast parallel computing environment with large memory (to store the ``non-bond list'' information). In particular, the proposed silicon graphics high-performance computer provides the ideal environment. In addition, advantage of the extremely efficient SGI fast-Fourier transform libraries can be used to compute the long-range electrostatics with linear-scaling Ewald methods [2, 19, 3]. Very rapid turn-around for these systems has been observed with excellent parallel scaling up to 16 processors (approximately 0.80 of the theoretical speed-up).

In the second stage of the NIH-supported research, a series of hybrid quantum mechanical/molecular mechanical (QM/MM) simulations [14] to characterize the details of the mechanism of the hammerhead ribozyme self-cleavage in the presence of divalent metal ions. This will be done using a new semi-empirical ``specific reaction parameter'' Hamiltonian especially designed for phosphate hydrolysis reactions (henceforth referred to as the SRP-PH Hamiltonian) [9]. The free-energy of the reaction will be calculated using umbrella sampling along an approximate reaction coordinate calculated using a conjugate peak refinement procedure [5]. Future work will also compare results with those from the replica-path method. These simulations are costly because of the consideration of a fairly large quantum region (one hundred atoms or more when fully coordinated divalent metals ions are included), and hence require considerably large memory. The SGI origin machines have traditionally been the most efficient for these calculations.

In the third stage of the NIH-supported research, a series of full electronic structure calculations [7, 16, 13] will be performed to take a more probing look at the quantum mechanical nature of the chemical reactions [17, 18, 20]. In the case of the model system, high-level density-functional calculations in the presence of fully coordinated metal ions and key solvent molecules are performed, in addition to semi-empirical calculations with the SRP-PH Hamiltonian for comparison.

Finally, linear-scaling semi-empirical calculations on the entire hammerhead ribozyme system will be performed [8] to study the reaction profile fully quantum mechanically. These calculations require a huge amount of memory, and parallelization is extremely efficient on the high-performance shared-memory SGI platforms.

In addition to the above summary of our computational efforts, there is a great need for our lab to have access to state-of-the-art biomolecular visualization and interactive graphical support. It is without question that the SGI systems are the premiere graphical machines in commercial and free visualization software and hardware.

The goal is to use these calculations to paint a more clear, broad picture of the structure, dynamics and reaction mechanism of this key prototype RNA enzyme. To accomplish this goal during the period of the NIH funding requires a great deal of computational support, beyond the current capability of the facilities at the MSI. The proposed computational equipment would greatly, if not altogether, relieve this resource bottleneck.